The Codex of A Companion of The Prophet and The Qurān of The Prophet

Arabica 57 (2010) 343-436 brill.
nl/arab
The Codex of a Companion of the Prophet and the

Qurʾān of the Prophet
Behnam Sadeghi and Uwe Bergmann1
Abstract
The essay discusses a manuscript of the Qurʾān dating from the first half of the seventh century
AD. The text does not belong to the ʿUt ̠mānic textual tradition, making this the only known
manuscript of a non-ʿUt ̱mānic text type. The essay compares this text type with those of
the ʿUt ̠mānic and other Companion textual traditions in order to shed light on the Prophetic
prototype.
Keywords
Qurʾān, Prophet Muḥammad, ʿUt ̱mān, Ibn Masʿūd, Islamic origins, palimpsest
1
We would like to thank the owner of the Stanford ’07 folio for making it available and for
submitting a sample for radiocarbon dating, Bryce Cronkite-Ratcliff for help with image trac-
ing, Ceci Evangelista of the Office of Development at Stanford University for logistical and
other help, and the staff at Stanford University Libraries and the SSRL. Portions of this research
were carried out at the Stanford Synchrotron Radiation Laboratory (SSRL) at the Stanford Lin-
ear Accelerator Center (SLAC), a national user facility operated by Stanford University on
behalf of the U.S. Department of Energy, Office of Basic Energy Sciences.
Behnam Sadeghi would like to thank the following persons for their written comments on
this essay: Michael Cook, David Powers, Patricia Crone, Michael Cooperson, Mohsen Goud-
arzi, Devin Stewart, Zaid Adhami, and Muhammad Mekki. I also thank Michael Cook and
Robert Waltz for helpful conversations and exchanges, Mette Korsholm for providing very
high-resolution images of the folio in the David Collection, M.S.M. Saifullah for bibliographi-
cal guidance, Scott Lucas for lending his UNESCO CD-ROM, and M.M. al-Aʿẓamī for gifting
the second edition of his History of the Qurʾānic Text. The first version of this essay was pre-
sented at the Colloquium on the Early History of the Qurʾān, Stanford University, July 30-31,
2009.
Uwe Bergmann’s contributions consisted of creating the instrumentation for and implement-
ing the X-ray fluorescence imaging of the folio called here Stanford ’07, determining the com-
position of the inks on the folio, and participating in preparations for radiocarbon dating.
Behnam Sadeghi converted the numerical output of imaging into images of the folio, traced the
lower text, calculated one-sided probabilities for the radiocarbon dating results, and wrote this
essay. The pronoun “I” in this essay refers to Behnam Sadeghi.
© Koninklijke Brill NV, Leiden, 2010 DOI: 10.1163/157005810X504518
Downloaded from Brill.com12/28/2021 08:48:21AM

via Fondation du Roi Abdul Aziz
344 B. Sadeghi, U. Bergmann / Arabica 57 (2010) 343-436
I. Introduction
Preview of Results
This is a study of two Qurʾāns: the upper and the lower layers of a palimpsest
called here Ṣanʿāʾ 1. The upper layer of writing, a standard Qurʾān, could be
from the first or second half of the seventh century AD, and possibly even
early eighth century. Radiocarbon dating assigns the parchment, and hence
the lower writing, to the first half of the seventh century.
Early Muslim reports assert that different Companions of the Prophet had
different versions of the Qurʾān, and some reports give the purported variants
of their codices. The differences among these codices appear to have moti-
vated an attempt at standardization. According to the collective memory of
early Muslims, the Companion ʿUt ̠mān, after becoming caliph, disseminated
a version of the holy book, declaring it the standard. The date of this event is
uncertain, but it appears to have taken place sometime during AH 24-30, i.e.
AD 644-650.2 It is to the textual tradition identified with this version that
almost all extant Qurʾānic manuscripts belong.
The main significance of the Ṣanʿāʾ 1 manuscript is that its lower text does
not belong to this ʿUt ̠mānic textual tradition. In this sense, it is “non-
ʿUt ̠mānic.” It belongs to some other textual tradition which is designated here
as C-1. The C-1 textual tradition is distinct not only from that of ʿUt ̠mān,
which is known from both literary sources and manuscripts, but also from
those of Companions Ibn Masʿūd and Ubayy b. Kaʿb, whose recensions of
the Qurʾān are not attested in manuscripts, being known only from descrip-
tions in literary sources. I will argue that C-1 and these others formed parallel
textual traditions. Comparing them can thus illuminate the state of the text
prior to the branching off of these various traditions. It can shed light on the
progenitor of all textual traditions, the Qurʾānic prototype.
The essay demonstrates that memory and orality played a role in the gene-
sis of the Companion codices. This fact fully accounts for the differences
between the textual traditions of C-1 and ʿUt ̠mān. The magnitude and num-
ber of the differences point to orality. Purely oral transmission, however, is
unlikely, since differences among the codices remain the exception rather
than the rule, even when it comes to minor elements of language. Nor does a
scenario of deliberate redaction of a written text explain the differences. The
2
For an analysis of the evidence bearing on the date of ʿUt ̠mān’s standardization, see Maḥmūd
Rāmyār, Tārīḫ-i Qurʾān, 2nd ed., Tehran, Amīr Kabīr, HS 1362/1983, p. 433-5; cf. Ahmad ʿAli
al-Imam, Variant Readings of the Qur’an, Herndon, Virginia, International Institute of Islamic
Thought, 2006, p. 20-1.

B. Sadeghi, U. Bergmann / Arabica 57 (2010) 343-436 345
evidence calls for a model involving both orality and writing, i.e. “semi-orality”.
In particular, a scenario in which the prototype was dictated provides the
appropriate model. The largest differences between the Qurʾānic textual tra-
ditions of C-1, ʿUt ̠mān, etc. go back to when the prototype was recited and
taken down by different scribes in somewhat different ways.
This is not to say that the lower writing of Ṣanʿāʾ 1 itself was the proximate
product of dictation. It may have been copied off a manuscript. The point is
that the lower writing and its parent manuscript were both part of a “C-1
textual tradition”. The C-1 textual tradition and the traditions corresponding
to other Companion codices must have branched off at some point. It is this
branching off that involved semi-orality. Once the various textual traditions
(the standard one and the Companion codices) were born through the dicta-
tion of some or all of them, they could have thereafter been handed down
through written transmission.
Textual criticism aims to determine a prototype on the basis of different
versions of a text. For the purposes of the present task, these versions include
the ʿUt ̠mānic Qurʾān and different Companion codices, and it is thus neces-
sary to discuss their status before undertaking the analysis. I begin with the
codex of ʿUt ̠mān. Drawing out the ramifications of important recent findings
(by Michael Cook, Yasin Dutton, Hossein Modarressi, etc.),3 I synthesize a
framework for conceptualizing the relationship of the ʿUt ̠mānic tradition
with non-ʿUt ̠mānic codices and for approaching ʿUt ̠mānic manuscripts such
as the upper text of the palimpsest. As for the status of non-ʿUt ̠mānic codices
described in Muslim literary sources, I argue that there must be a significant
kernel of truth in these reports, and that a particular second-century list of
the variants of the Companion Ibn Masʿūd may be a largely reliable represen-
tation of an early codex.
The first method of textual criticism used is stemmatics, sometimes called
the “genealogical method”, a procedure based on how frequently each text
agrees with each of the others. This approach is applied to three text types—
those of ʿUt ̠mān, C-1, and Ibn Masʿūd—to construct possible stemmata. The
most plausible stemma is shown to be one in which a prototype is the com-
mon ancestor of these three Companion codices. Subsequently, the data sug-
gests that among them, the textual tradition of ʿUt ̠mān gives the most
accurate reproduction of the prototype. An alternative scenario explains the
data equally well: one may envision ʿUt ̠mān’s codex as a composite formed by
3
The citations to the works of these scholars and others will follow in this essay in due
course. Incidentally, some authors have missed the importance of these works. See e.g. a num-
ber of papers in G.S. Reynolds (ed.), The Qurʾān in its Historical Context, London, Routledge,
2008.

comparing a variety of Companion codices and picking the most common

version where they differed. Stemmatics does not depend on the nature and
form of the variants or the character of a codex as a whole.
The nature of the differences is taken up in my second approach to textual
criticism, polarity analysis of internal evidence: I analyze two texts (the codi-
ces of C-1 and ʿUt ̠mān) to identify statistical asymmetries between them, and
determine which text type is likely to be earlier. The inference from asymme-
try to relative dating in this type of analysis follows principles that reflect the
mode of transmission. The Companion codices demand principles that are
suited to cases of concomitant written and oral transmission, as in dictation.
The essay expounds such principles and applies them to the differences
between the codices of ʿUt ̠mān and C-1. This approach, too, indicates that
the wording of ʿUt ̠mān’s codex is older than that of C-1.
The analysis paves the way for a discussion of the ancestor of all Qurʾānic
textual traditions, the “prototype.” The evidence analyzed in this essay indi-
cates that the prototype is to be identified with the Prophet Muḥammad.
Aside from the issue of dating, the evidence allows a discussion of other char-
acteristics of the prototype. The conclusions here diverge from the usual pre-
modern, traditional account towards a more conservative understanding. The
most secure conclusion of the present study is that the sequences of verses
and sentences were fixed already in the Prophetic prototype. This result is a
departure from mainstream traditional views about the date when the pieces
of revelation were joined together into fixed sūras. It is traditionally more
common to date this to the reign of ʿUt ̠mān. The findings here support the
less common dating to the time of the Prophet.
Another way in which the conclusions here differ from traditional views
concerns the relative merits of the different Companion codices. The major-
ity traditional scholarly position associates Companion Qurʾāns with the
Seven Modes (aḥruf ) in which God revealed the holy book. It thus does not
claim for the ʿUt ̠mānic Qurʾān greater verbal fidelity to the original than the
codices of Companions such as Ibn Masʿūd or, by implication, C-1. (Never-
theless, the majority position considers ʿUt ̠mān’s version as the only one fit
for recitation in worship. This is justified by citing consensus or by calling it
the Prophet’s last version, i.e. Muḥammad’s last presentation of the Qurʾān
to, or from, Gabriel.)4 On the other hand, an early minority strand of opin-
4
On the majority pre-modern position on Companion codices, see Ibn al-Ǧ azarī, Taqrīb
al-našr fī l-qirāʾāt al-ʿašr, Cairo, Muṣtạ fā l-Bābī, 1381/1961, p. 32-4; cf. Ibn ʿAṭiyya, ʿAbd
al-Ḥ aqq b. Ġ ālib b. ʿAbd al-Raḥmān, al-Muḥarrar al-wağīz fī tafsīr al-kitāb al-ʿazīz, Beirut, Dār
al-kutub al-ʿilmiyya, 1413/1993, p. 1-49. Ibn al-Ǧ azarī writes that in his time important and
unimportant scholars alike believed that seven modes (aḥruf ) in which the Qurʾān was revealed

ion, espousing a kind of codical pluralism, considered non-ʿUt ̠mānic Qurʾāns

as valid in all respects. Neither approach took the standard Qurʾān to be a
more accurate representation of the Prophet’s recitation. The present essay,
however, finds that the textual evidence supports a more conservative conclu-
sion, namely that the ʿUt ̠mānic tradition is better than C-1 in reproducing
the Prophetic prototype. (This conclusion, however, may be modified in
future if the parts of the manuscript not studied here do not support the
trends observed thus far.)
When thinking about the text of an individual manuscript, it is necessary
to understand it as a representative of a text type. The manuscripts belonging
to the standard textual tradition, i.e. the ʿUt ̠mānic tradition, may have many
differences among themselves; but their differences are nonetheless minor
when these manuscripts are compared to the C-1 or Ibn Masʿūd texts. We
may therefore speak of an ʿUt ̠mānic text type, of which each individual
ʿUt ̠mānic manuscript is an instance despite the differences it may exhibit
with other ʿUt ̠mānic texts. A text type is thus a cluster of similar texts which
stand apart from other clusters. The text type is an essential notion for textual
criticism, for behind it lies the concept of a textual tradition. Texts that are in
the same text type are generally more closely related to one another in terms
of origination. They are likely to belong to the same branch of the textual
tradition. Text types thus allow one to arrange texts in a family tree just as
similar phenotypes signify evolutionarily related species of animals and plants.
It will be convenient to refer to the manuscript studied here, i.e. to both of
its layers, and to parts of it that are not extant, as Ṣanʿāʾ 1. The Companion of
the lower codex, whoever he may have been, the codex of this Companion,
and its corresponding text type and textual tradition, will be referred to as
C-1, standing for “Companion 1”. Context will disambiguate the sense. Sim-
ilarly, the names ʿUt ̠mān and Ibn Masʿūd will be applied equally to the per-
sons, the codices associated with them, and the resulting text types.
It should be noted that, throughout this essay, by “ ʿUt ̠mānic”, I do not
mean “adhering to the ʿUt ̠mānic orthography”, which is the sense in which
some writers use this adjective. Rather, I mean “belonging to the text type of
the standard Qurʾān”. Standing behind this text type is the textual tradition
of the standard Qurʾān—a tradition that reportedly began with the codices
ʿUt ̠mān disseminated as the ancestors of all the manuscripts in the textual
tradition. In the course of dozens or hundreds of copyings intervening since
went well beyond the ʿUt ̠mānic text. He quotes one authority to the effect that other modes can
include additions, omissions, substitutions, and transpositions of words. These are precisely the
types of variants found in C-1 and the reported codex of Ibn Masʿūd. One may add that some
pre-modern tafsīrs quoted the codex of Ibn Masʿūd.

the time of the ancestor manuscripts, many small changes may have accumu-
lated. By far, the most common type of change concerns spelling. The spell-
ings of Arabic words were not standardized and an early scribe often followed
his own discretion rather than the conventions of the manuscript he was copy-
ing. Thus, an ʿUt ̠mānic manuscript, in the sense I use the term, needs not
preserve the original orthography of its ancestor manuscript. It may also dif-
fer from the ancestor manuscripts in ways that are more significant.
Ṣanʿāʾ 1 Folios Identified and the Results of Radiocarbon Dating Described
The research reported here began with the examination of a single folio of a
palimpsest, called here “Stanford ’07”, of dimensions 36.3 cm by 28.5 cm.
The study then extended to photographs of some of the other leaves from the
same manuscript. A palimpsest is a parchment with at least two layers of text.
The older layer is scraped off or washed to make room for the new text, but
over time it resurfaces as a shadow, in this case as a pale brown text. In August
2007, Uwe Bergmann subjected the Stanford ’07 folio to X-Ray fluorescence
imaging at the Stanford Synchrotron Radiation Laboratory in the Stanford
Linear Accelerator Center (SLAC). X-Ray fluorescence imaging is a tech-
nique for detecting, and tracing, the chemicals left on a leaf by inks or dyes.
Its application to the Stanford ’07 folio assisted in reading and tracing out
the lower text, bringing to light some letters, verse separators, and diacritical
marks not otherwise visible or legible (Figures 3-4). Moreover, the fact that
the inks used in the two layers were chemically different made it possible to
determine to which layer every feature belongs. For example, it was possible
to confirm that the diacritical marks and verse dividers were in the same ink
as the main text. This holds true for both the lower and upper texts. The
same is true of the decorative sūra separators of the lower text. Therefore,
these features probably were not added at a later stage. High-resolution pho-
tographs of the folio, images of the traces of chemicals, and a hand-drawn
tracing of the lower text can be downloaded from http://ssrl.slac.stanford.edu/
quranleaf.
Radiocarbon dating was performed on a sample of Stanford ’07. The anal-
ysis was done at the Accelerator Mass Spectrometry (AMS) Laboratory at the
University of Arizona. The results indicate that the parchment has a 68%
probability of belonging to the period between AD 614 to AD 656. It has a
95% probability of belonging to the period between AD 578 and AD 669.5
5
The results of radiocarbon dating at the NSF Arizona AMS Laboratory at the University of
Arizona were described in a letter dated May 23, 2008, by the director of the lab and Professor
of Geosciences and Physics, A.J. Timothy Jull. What follows is a quotation from the letter:

Figure 1. Stanford ’07, recto.

Figure 2. Stanford ’07, verso.

Figure 3. Stanford ’07, lower text, recto. Kor 2, 191-6; or by the folio’s own
numbering, Kor 2, 188-93.

Figure 4. Stanford ’07, lower text, verso. Kor 2, 197-205; or by the folio’s
own numbering, Kor 2, 194-202.

For historical reasons, however, what is of greater interest is the probability

that the parchment is older than a certain date. Such probabilities are readily
calculated and are shown in Table 1.6 The probability that the parchment
is older than AD 646 is 75.1 %, or a three-to-one likelihood. It is highly
probable therefore, that the Ṣanʿāʾ 1 manuscript was produced no more than
15 years after the death of the Prophet Muḥammad.
Table 1. Result of Radiocarbon Dating. The graph gives, in five-year

increments, the probability (%) that the parchment is older than a certain
date AD, while the table to the right gives the probabilities for selected
years.
Probability that the

parchment is older
Date
Probability (%)
(AD)
100
90
80
70
60 635.5 56.2 %
50 645.5 75.1 %
655.5 91.8 %
40
660.5 95.5 %
30
670.5 98.8 %
20 675.5 99.2 %
10
0
550 570 590 610 630 650 670
“δ13C −20.8 per mil

Fraction modern 0.8393 ± 0.0037
Radiocarbon Age 1,407 ± 36 years BP
Calibrated Age 614-656 AD (1 sigma, 68% confidence)
578-669 AD (2 sigma, 95% confidence)
The 14C measurements were done by accelerator mass spectroscopy (AMS) and I have also
quoted the stable-isotope value of 13C/12C, given in δ13C units. The values are quoted corrected
to −25% for δ13C, which indicates the value of stable-isotope ratio of 13C/12C deviation from a
known standard, in parts per mil (‰). Most organic materials are typically about −25‰.
The radiocarbon age is the conventional 14C age and is quoted in years ‘before present,’ where
‘present’ has been defined as the expected natural level for ~1950 AD. As the radiocarbon con-
tent of the atmosphere fluctuates with time, it is necessary to calibrate this value against known
material, usually known-age tree rings.”
6
I obtained the probability distribution from the radiocarbon age 1,407 ± 36 years BP using
the IntCal 04 calibration and the software on the website of the Oxford Radiocarbon

The date of the parchment is a reliable indicator of the date of the lower writ-
ing. The parchment probably is not many years older than the lower writing.
Given its dimensions, this manuscript must have been expensive, requiring a
whole flock of animals. It is unlikely that the folios required for this Qurʾān
were procured for a purpose other than the one to which they were put. In
the initial decades of Islam, the period to which this manuscript belongs, the
Arabs did not have many books to copy beside the Qurʾān. Indeed, the only
extant vellum manuscripts of a comparable size in the Ḥ iǧāzī script are with-
out exception Qurʾāns. There would not have been a large supply of unused
folios of this size.
There is no public record of the Stanford ’07 folio prior to its being auc-
tioned by Sotheby’s in London in 1993. At least three other folios from Ṣanʿāʾ
1 have been sold in auction houses in London, and images of these have been
used in the present study. A leaf that would have been adjacent to Stanford
’07 in the codex was auctioned by Sotheby’s in 1992, resurfaced in a Chris-
tie’s auction on May 1, 2001, emerged in the Sam Fogg Gallery,7 and finally
moved to the David Collection in Copenhagen, where it now resides. A leaf
was auctioned by Bonhams in 2000, and another by Christie’s on April 8,
2008.8 9 At least thirty-two other leaves from Ṣanʿāʾ 1 are in Ṣanʿāʾ, Yemen, in
the collection of manuscripts retrieved from the Grand Mosque of Ṣanʿāʾ,
bearing the catalog number 01-27.01.
Table 2 identifies the folios included in this study. The quality of the avail-
able images varies. For example, for the Bonhams folio I was limited to the
image in the printed auction house catalog, while the David Collection
kindly provided very high-resolution photographs. Of the leaves in Ṣanʿāʾ, a
few low-resolution images have been published, useful only for the study of
the upper text. My treatment of the lower text, therefore, is limited to the
four folios sold in London.
The upper writing is the work of two scribes. One scribe wrote the first
two folios (from the end of sūra 2), and another scribe wrote the others. Each
Accelerator Unit (https://c14.arch.ox.ac.uk/login/login.php?Location=/oxcal/OxCal.html). The

results are consistent with the calibrated age announced by Timothy Jull. I subsequently calcu-
lated running totals to obtain the probabilities for one-sided intervals shown in Table 1.
7
Marcus Fraser and Will Kwiatkowski, Ink and Gold: Islamic Calligraphy, London, Paul
Holberton, 2006.
8
The Stanford ’07 folio was first disclosed in Sotheby’s, Oriental Manuscripts and Miniatures,
1993, p. 18-23. For the other leaves see e.g. Sotheby’s, Oriental Manuscripts and Miniatures,
1992, p. 254-9; Bonhams Knightsbridge, Islamic and Indian Works of Art, 2000, p. 11-4; Chris-
tie’s, Art of the Islamic and Indian Worlds, 2008, p. 24-7.
9
The Christie’s folio was sold for $4,924,279, breaking the world auction record for any
Islamic manuscript (http://www.christies.com/presscenter/pdf/04082008/144844.pdf, accessed
in August 2008).

Table 2. The folios of Ṣanʿāʾ 1 studied here.
Name of Leaf and Upper Text Lower Text

Source
Stanford ’07 Recto: Recto: 2, 191 (first word) to near
(Originally 2, 265-271 the end of 2, 196.
Sotheby’s ’93) Verso: Verso: 2, 197 (ašhurun)—2, 205
2, 271-277 (the last fully visible word is yuh-
lika, last inferred word al-fasāda).
David Collection Recto: Recto: 2, 206 (wa-id̠ā)—2, 217
(originally 2, 277-282 (the last visible word is wa-l-
Sotheby’s ’92), Verso: masǧidi, and the last inferred
Copenhagen, inv. 282-286 word minhu).
no. 86/2003. Verso: 2, 217 (akbaru ʿinda
llāhi)—2, 223 (ḥ art̠un lakum).
Bonhams 2000 Recto: 4, 33-43 Recto: 5, 41 (yuṭahhira)—5, 47
Verso: 4, 43-56 ( fa-ulāʾika hum).
Verso: 5, 48 (. . . li-mā bayna
yadayhi min al-kitābi muhaymi-
nan; this last is the first mostly
visible word)—5, 54 (llāhi wa-lā).
Christie’s 2008 Recto: 4, 171-5, 3 Recto: Beginning of 63 (bi-smi
Verso: 5, 3-9 llāhi l-raḥ māni l-raḥ īmi) to the
end the sūra; the beginning of 62,
with the basmala, to 62, 11
(tiǧāratan aw lahwan).
Verso: 62, 11 (infaḍḍū) to the end
of the sūra; entire sūra 89 with the
basmala; sūra 90 (with the bas-
mala) to 90, 6 (yaqūlu).
Noseda 1a 6, 49—6, 61 The image available to the author
UNESCO Image is not legible.
152255B
152254B

Table 2 (cont.)
Name of Leaf and Upper Text Lower Text

Source
Noseda 4a 16, 73-89 The image available to the author
027002B
Noseda 5a 16, 102-118 The image available to the author
027004B
152256B
Noseda 14b 28, 74—28, 86 15, 54-72
Maṣāḥif Ṣanʿāʾ,
1985, inside
cover, partial
image.
Noseda 17a 31, 24—32, 4 Possibly 9, 103-113. The image
Maṣāḥif Ṣanʿāʾ, available to the author is not
1985 legible.
Hans-Casper von (a bifolio) is not legible.
Bothmer, 1987
152257B
had a distinct handwriting and distinctive dotting and orthographical habits.
(‫)ﻋﻠﻰ‬, whereas the second one dotted the yāʾs and spelled ʿalā with alif (‫)ﻋﻼ‬.10
For example, the first scribe tended to dot the tāʾs and spelled ʿalā with yāʾ
10
François Déroche has noted, “in manuscripts written by two or more copyists, it is not
infrequent to notice that each of them had his own way of putting diacritical dots, with regard
either to their density, or to the individual letters receiving the dots” (François Déroche, “New
Evidence about Umayyad Bookhands”, in Essays in Honour of Salah al-Din al-Munajjid, Lon-
don, al-Furqan Islamic Heritage Foundation, 2002, p. 611-42, 627).

Table 3. The four hands in Ṣanʿāʾ 1, listed in the proposed chronological

order. This essay is concerned exclusively with the first two.
Name of Hand Prevalence Appearance Iron & Copper in Ink
Lower Text systematic Light brown Yes
Upper Text systematic Dark brown Yes
Upper Modifier occasional Black, poor hand No
Lower Modifier occasional Black, excellent No
hand
Both literary and paleographic evidence show that there was nothing unusual
about two or more persons collaborating to copy a Qurʾān.11 As for the lower
text, it appears to be entirely in the hand of one scribe, although a closer
examination is needed to verify this. Interestingly, the first two folios listed in
Table 2 are adjacent in both the lower and upper layers.
Basic Description of the Folios
There are four layers of writing on Ṣanʿāʾ 1. These are shown in Table 3. The
“lower text” and “upper text” apparently constituted complete codices of the
Qurʾān. The upper one is ʿUt ̠mānic and the lower one is not. The “upper
modifier” (in the David and Stanford ’07 leaves) appears as an awkward,
amateurish hand that wrote over the words where the upper text had faded,
especially near the edges. The hand designated as “lower modifier” (in the
David and Stanford ’07 leaves) did two things. First, it wrote over some
words and letters of the lower text that had faded or had not resurfaced
because the ink had been thoroughly removed by the original scraping off of
the lower writing.12 Second, on occasion it wrote the standard version where
11
For literary evidence confirming that there could be collaboration in copying, see the report
of ʿAlī b. al-Ḥ usayn b. ʿAlī (d. 94/712), of which there are at least two versions related through
his grandson, Ǧ aʿfar, as the common link: Abū Bakr ʿAbd Allāh b. Abī Dāwūd Sulaymān
b. al-Ašʿat ̠ al-Siǧistānī, Kitāb al-Maṣāḥ if, ed. Arthur Jeffery, Egypt, Maṭbaʿat al-Raḥmāniyya,
1355/1936, p. 166; Abū Bakr Aḥmad b. al- Ḥ usayn b. ʿAlī l-Bayhaqī, al-Sunan al-kubrā, Dār
al-fikr, Beirut, n.d., VI, p. 16. The paleographic evidence consists of early manuscripts that are
in more than one hand. Such manuscripts are invaluable for the paleographer, for they give evi-
dence of the range of variation in calligraphic and orthographic practices in a single time and
place. See François Déroche, “New Evidence about Umayyad Bookhands”, p. 611-42, 629; id., La
transmission écrite du Coran dans les débuts de l’islam: Le codex Parisino-petropolitanus, Leiden,
Brill, 2009, p. 26-77; Intisar Rabb, “Non-Canonical Readings of the Qurʾān: Recognition and
authenticity (The Ḥ imṣī reading)”, Journal of Qurʾānic Studies, 8/2 (2006), p. 84-127.
12
Here is an example of the lower modifier writing over a faded or damaged part of the

the lower text departed from it. However, it does neither of these things sys-
tematically. And it does not try to obscure the deviations from the standard
version. The evidence shows that the lower modifier came after the erased
lower text had reemerged, hence after the upper text.13 It could even be an
early modern intervention. So, the lower and upper texts are the earliest ones.
The question arises which script came earlier: the upper-modifier or the low-
er-modifier. The upper-modifier seems to have been added to ensure the con-
tinued usability of the codex despite the faded words. The lower-modifier
seems to have been added at a time when usability was no longer a concern. I
conjecture, therefore, that it came last. A thorough physical examination is
needed, however, before the chronological order of the last two writings is
settled beyond doubt. In any case, among the four layers, the upper and
lower texts are the oldest, while the “modifier” texts, having come later, hold
relatively little interest. It is thus with the lower text and upper text that this
essay is exclusively concerned.
A few noteworthy differences between the upper and lower writings may
be mentioned. The upper script is more generous than the lower one with the
tiny dashes that distinguish consonants of a similar shape. The fact that the
lower script exhibits such marks is important; but it is not wholly unexpected,
since a commercial papyrus from the year 22/642 also has such marks.14
So does an inscription from the year 24/644-5,15 in addition to other early
documentary sources.16 Moreover, pairs of such dashes are arranged, more
lower text: most of the mīm of al-ḫiṣāmi in Kor 2, 204 had been scraped off, and no trace of the
ink is detectable by X-Ray Fluorescence Imaging. Nonetheless, we know that there was a mīm
there because the telltale tip of its tail survived damage. The lower modifier wrote a full mīm in
the expected place.
13
The strongest indication that the lower modifier came after the upper text is that it was
not erased before the upper text was written. Another indication that the lower modifier came
after both the lower and upper writings is that its ink is chemically unlike theirs and similar to
that of the upper modifier. It contains no iron or copper and is black, being apparently carbon-
based. By contrast, the inks of the lower and upper writings both contain iron and copper.
Finally, the calligraphic style of the lower modifier suggests that it did not belong to the first
two centuries AH. The lower modifier gives one the impression of somebody who read the
emerged lower text out of curiosity and casually noted some discrepancies with what he remem-
bered of the Qurʾān.
14
Alan Jones, “The Dotting of a Script and the Dating of an Era: The strange neglect of
PERF 558”, Islamic Culture, 72 (1998), p. 95-103.
15
ʿAli ibn Ibrahim Ghabban, Robert Hoyland, “The inscription of Zuhayr, the oldest Islamic
inscription (24 AH/AD 644-645), the rise of the Arabic script and the nature of the early Islamic
state”, Arabian Archaelogy and Epigraphy, 19/2 (2008), p. 210-37.
16
There are even pre-Islamic inscriptions with diacritical marks distinguishing consonants.
For a non-exhaustive list and images of some of the early sources with such marks see
Muḥammad Muṣtạ fā l-Aʿẓamī, The History of the Qurʾānic Text, 2nd ed., Riyadh, Azami Pub-
lishing House, 2008, p. 152-6.

often than not, with an upward slope in the upper script and a downward
slope in the lower one. The upper writing has more words per page than the
lower one. The upper layer uses a special marker for every ten verses. By con-
trast, the lower text marks the 200th verse of sūra 2 by means of what looks
like the letter wāw encircled by short dashes. The symbol is similar to those
used in the BNF arabe 328 (a) manuscript to mark groups of fifty verses (i.e.
the 50th, 100th, and 150th verses).17 The chemical difference between the
inks used has already been noted.
There are a number of ways in which the lower text is more “elaborate”
than the upper one, thus complicating the assumption that more sophisti-
cated means later. The lower text has decorations between sūras, but the
upper text does not.18 The lower text inserts captions between sūras such as,
“This is the end of sūrat al-Munāfiqīn”; but the upper text lacks sūra names.
There is resonance here with literary sources: certain first-century figures such
as Ibrāhim al-Naḫaʿī (d. 96/715, Kūfa) and ʿAṭāʾ (d. 114/732, Mecca) were
opposed to captions that said, “This is the end of sūra such and such”, or
stated the number of verses in the sūra.19
Surprisingly, the lower script on occasion appears to use what are possibly
diacritics, in the form of perfectly round dots, to signify short vowel marks
(and possibly elided alifs, i.e. hamzat al-waṣl). These dots are in the same ink
as the rest of the lower writing and do not appear to have been added later.
In particular, the words tahluka and nusukin in Stanford ’07, in Kor 2, 195
and 196, appear to have dots on the consonants lām (“l”) and sīn (“s”) respec-
tively, both placed at an elevated level, possibly for indicating the vowel “u”
(the ḍamma). The vocalization of both of these letters later became a matter
of dispute. So, it seems unlikely that the dots are in these positions by chance.
If these words are vowelled, this may indicate that they were already deemed
difficult at the time of C-1. In each case, the lower writing’s apparent vowel-
ling agrees with what came to be the majority view of later authorities.20
17
See François Déroche and Sergio Noseda, Sources de la transmission manuscrite du texte
coranique: I Les manuscrits de style ḥ iǧāzī, Volume i, Le manuscrit arabe 328 (a) de la Bibliothèque
nationale de France, Lesa, Fondazione Ferni Noja Noseda, 1998, p. 29 (Kor 3, 100), p. 107
(Kor 6, 50), p. 116 (Kor 6, 100), p. 155 (Kor 7, 150).
18
The fact that the lower text sometimes has decorative markings between sūras has already
been noted by Sheila Blair, Islamic Calligraphy, Edinburgh, Edinburgh University Press, 2006,
p. 119, with regard to one of the folios located in Ṣanʿāʾ, namely Noseda 18a. Such sūra divid-
ers are found also in the Christie’s folio, but not some other folios.
19
Ibn Abī Dāwūd, al-Maṣāḥ if, p. 137-41.
20
For the disagreements over the vowelling of these words, see ʿAbd al-Laṭīf Muḥammad
Ḫ at ̣īb, Muʿǧam al-qirāʾāt, Damascus, Dār Saʿd al-Dīn, 2002, I, p. 266 and 269.

However, more folios need to be analyzed before one states with certitude
that the dots are indeed diacritical marks, as opposed to accidental ink
droplets.
Apart from these distinctions, the upper and lower scripts have fairly simi-
lar calligraphic styles.21 They also use a similar end-of-verse marker in the
form of several short dashes; but in the upper layer these are often arranged
in triangles. Both writings lack additional spacing between words; a word
may be split at the end of the line; the number of lines can vary from one
page to the next; and the lines of text are not always strictly straight or per-
spelling ‫( ﺍﻭﻻ ﺍﻻﻟﺒﺐ‬in the accusative) instead of ‫ ﺍﻭﻟﻰ ﺍﻻﻟﺒﺐ‬. This spelling is
fectly horizontal.22 Another shared feature of the two layers is the distinctive
attested in several other Qurʾāns in the Ḥ iǧāzī script.23

The non-ʿUt ̠mānic status of the lower text is shown by two things. First, its
verbal differences with the ʿUt ̠mānic Qurʾān show that it represents a differ-
ent text type and may therefore belong to a different branch of the family
tree. Moreover, its variants are similar in nature to those reported of Com-
panion codices. The second indication is its sūra ordering. On the Christie’s
folio, sūras 63, 62, 89, and 90 appear in that order, a sequence that is very
close to what has been reported for the codex of Ubayy b. Kaʿb, a point that
will be discussed further (p. 393). In terms of wording, the lower text also
agrees with reported non-ʿUt ̠mānic variants in a few cases, as shown in Table
4; however, as a rule, reported non-ʿUt ̠mānic variants do not appear in C-1,
nor are the variants of C-1 reported in the sources. Thus C-1 should not be
identified with the codices whose variants have been described in the literary
sources (Ibn Masʿūd or Ubayy b. Kaʿb); it represents an independent codex,
text type, and textual tradition.
21
Based on a preliminary examination of the lower and upper scripts, there appears to be no
letter shape in one script that is not attested in the other, although the two scripts may use the
different shapes of the same letter with different frequencies. In the upper writing, the letters
are more angular; for example, the nūns sometimes display sharp turns.
22
This last point, the lack of linear discipline, has been noted already in Sotheby’s, Oriental
Manuscripts and Miniatures, 1993, p. 18-23, 22.
23
In some other early Qurʾāns, the term ūlū l-albābi is spelled with an alif in the nominative
and accusative cases and with a yāʾ in the genitive. See e.g. Maṣāḥ if Ṣanʿāʾ, Dār al-āt ̠ār
al-islāmiyya, Kuwait National Museum, 1405/1985, no. 11 (catalog number 00-29.1). The MS
arabe 328 (a) in the Bibliothèque nationale de France has an alif in Kor 3, 7 (p. 18), Kor 13, 19
(p. 213), and Kor 14, 52 (p. 225). It has a yāʾ in Kor 3, 190 (p. 6) and Kor 12, 111 (p. 209).
The British Library’s Or. 2165 has an alif in Kor 13, 19 (p. 123) and Kor 14, 52 (p. 136). It has
a yāʾ in Kor 12, 111 (p. 119). See Déroche, Sources . . . arabe 328(a); id., Sources de la transmis-
sion manuscrite du texte coranique: I Les manuscrits de style ḥ iǧāzī, Volume ii, Le manuscrit Or.
2165 (f. 1 à 61) de la British Library, Lesa, Fondazione Ferni Noja Noseda, 2001.

Table 4. Instances where C-1 agrees with a reported non-ʿUt̠mānic

variant. Words that are only partially legible are in parantheses.24
Verse ʿUt ̠mān C-1 Codex agreeing with C-1
2.217 qitālin fīhi (wa)-ʿan qitālin fīhi ʿan qitālin fīhi

“They ask you about “They ask you about “They ask you about
the holy month— the holy month and the holy month, about
fighting in it.”25 about fighting in it.” fighting in it.”
(Reading of Ibn
Masʿūd.)26
2.222 fa-ʿtazilū l-nisāʾa ( fa-lā taqrabū) l-nisāʾa wa-lā taqrabū l-nisāʾa
fī l-maḥ īḍi wa-lā fī (maḥ īḍihinna) fī maḥ īḍihinna wa-
taqrabūhunna ḥ attā ḥ attā yataṭahharna ʿtazilūhunna ḥ attā
yaṭhurna yataṭahharna
“Avoid sexual relations “Do not go to “Do not go to
with women during women during their women during their
menstruation and do menstruation until they menstruation and avoid
not go to them until are clean” sexual relations with
they are cleansed” them until they are
clean”
(The reading of Ibn
Masʿūd and the Codex
of Anas. Both Ubayy
b. Kaʿb & Ibn Masʿūd
preferred yataṭahharna.)27
24
The first three cases in the table have already been mentioned by Alba Fedeli, “Early
Evidences of Variant Readings in Qurʾānic Manuscripts”, in Die dunklen Anfänge: Neue For-
schungen zur Entstehung und frühen Geschichte des Islam, K.-H. Ohlig & G. Puin (eds), Berlin,
Verlag Hans Schiler, 2007, p. 293-316.
25
My translations of Qurʾānic passages are often based on those of other translators. I have
consulted the following: M.A.S. Abdel Haleem, The Qurʾān: A New Translation, Oxford, Oxford
University Press, 2004; Arthur Arberry, The Koran Interpreted, London, Oxford University Press,
1983; Muhammad Asad, The Message of the Qurʾān, Gibraltar-London, Dar al-Andalus, 1980;
Abdalhaqq Bewley, Aisha Bewley, The Noble Qurʾān: A New Rendering of its Meaning in English,
Norwich, Bookwork, 1999; Thomas Cleary, The Qurʾān: A New Translation, Chicago, Starlatch
Press, 2004; E.H. Palmer, The Qurʾān: Translated by E.H. Palmer, Oxford, The Clarendon Press,
1900; Marmaduke Pickthall, The Meaning of the Glorious Qurʾān: Text and Explanatory Transla-
tion, Beirut, Dār al-kitāb al-lubnānī, 1971; J.M. Rodwell, The Koran, London, Dent, 1978;
George Sale, The Koran, New York, Garland, 1984; Muḥammad Saʿīd Šakīr, The Glorious Qurʾān,
Qum, Anṣāriyān, 1998; Mawlawi Sher Ali, The Holy Qurʾān: Arabic Text and English Translation,
14th ed., Surrey, Islam International, 1989; ʿAbdullāh Yūsuf ʿAlī, The Meaning of the Holy Qurʾān,
Brentwood, Md., Amana, 1991.
26
Ḫ atị̄ b, Muʿǧam al-qirāʾāt, I, p. 298.
27
Ḫ atị̄ b, Muʿǧam al-qirāʾāt, I, p. 308-9.

Table 4 (cont.)
Verse ʿUt ̠mān C-1 Codex agreeing with C-1
5, 45 wa-katabnā ʿalayhim wa-katabnā ʿalā banī Either wa-anzala llāhu
Isrāʾīla ʿalā banī Isrāʾīla or wa-
anzalnā ʿalā banī Isrāʾīla.
“We prescribed for “We prescribed for the Either “God sent down
them” Children of Israel” upon the Children of
Israel” or “We sent down
upon the Children of
Israel” (Codex of Ubayy
b. Kaʿb)28
63, 7 yanfaḍḍū yanfaḍḍū min ḥ awlihi The quotation of the
verse in a ḥ adīt̠ is
identical to C-1.
“that they disperse” “that they disperse from
around him”
The last item in Table 4 (viz. Kor 63, 7) is fascinating because here C-1
matches an unidentified codex that is quoted in passing in a ḥ adīt̠, a tradition
of Zayd b. Arqam. Five different transmitters related the tradition through
this chain of transmitters: Isrāʾīl [b. Yūnus b. Abī Isḥāq] (d. AH 160-2)—Abū
Isḥāq [al-Sabīʿī, ʿAmr b. ʿAbd Allāh] (d. 127, Kūfa)—Zayd b. Arqam (d. 65-8,
Kūfa)—Prophet. In this report, Zayd discusses the circumstances surround-
ing the Prophet’s dissemination of Kor 63, 7, and in doing so he quotes the
verse. His quotation contains the phrase min ḥ awlihi, which is found in C-1
but not in the standard Qurʾān.29 This means that one of the persons in the
above chain of transmission knew the verse with this extra element.
28
Ḫ atị̄ b, Muʿǧam al-qirāʾāt, II, p. 279.
29
Ibn Ḥ anbal, Abū ʿAbd Allāh Aḥmad b. Muḥammad, al-Musnad, Lebanon, Dār Ṣādir,
IV, p. 373; al-Buḫārī, Abū ʿAbd Allāh Muḥammad b. Ismāʿīl, Ṣaḥ īḥ al-Buḫārī, n.p., Dār al-fikr,
1401/1981, VI, p. 63-5; Abū l-Ḥ usayn Muslim b. Ḥ aǧǧāǧ, Ṣaḥ īḥ Muslim, Beirut, Dār al-fikr,
VIII, p. 119; al-Tirmid̠ī, Abū ʿĪsā Muḥammad b. ʿĪsā, Sunan al-Tirmid̠ī, 2nd ed., n.p., Dār al-
fikr, 1403/1983, V, p. 88; al-Nasāʾī, Aḥmad b. ʿAlī b. Šuʿayb, al-Sunan al-kubrā, Beirut, Dār al-
kutub al-ʿilmiyya, 1411/1991, VI, p. 492; al-Ḥ ākim al-Nīsābūrī, al-Mustadrak ʿalā l-Ṣaḥ īḥ ayn,
Dār al-maʿrifa, Beirut, II, p. 488; al-Ṭabarānī, Sulaymān b. Aḥmad b. Ayyūb Abū l-Qāsim,
al-Muʿǧam al-kabīr, 2nd ed., ed. Ḥ amdī ʿAbd al-Maǧīd, n.p., Dār iḥyāʾ al-turāt ̠ al-ʿarabī, V,
p. 186, 189; cf. Ibn Ḥ aǧar al-ʿAsqalānī, Aḥmad b. ʿAlī, Fatḥ al-Bārī, 2nd edition, Beirut, Dār
al-maʿrifa, VIII, p. 494, where he rejects the view of those who inferred from this ḥ adīt̠ that the
phrase min ḥ awlihi had been part of a codex.

The non-ʿUt ̠mānic status of the lower text has been noted already by Ser-
gio Noseda, Yasin Dutton, and Alba Fedeli.30 In 2007, Fedeli published a
report on two of the folios (David and Bonhams). She mentions a number of
important variants and points out the first three items in Table 4. Aside from
that, her discussion is slightly imprecise.31 Moreover, she misreads the text at
one point, mistakenly inferring an omission in the Bonhams folio, and then
uses that to argue that the version in the ʿUt ̠mānic Qurʾān resulted from
deliberate addition motivated by religio-political concerns.32
The upper text is definitely ʿUt ̠mānic. It exhibits the kinds and magnitude
of deviation from the standard text that typically accumulated in the course
of written transmission. It is to the ʿUt ̠mānic textual tradition and the place of
the upper text in it that I now turn.
30
Sergio Noseda, “La Mia Visita a Sanaa e il Corano Palinsesto”, Istituto Lombardo (Rendi-
conti Lett.), 137 (2003), p. 43-60; Anonymous, “‘The Qur’an: Text, interpretation and transla-
tion’ 3rd Biannual SOAS Conference, October 16-17, 2003”, Journal of Qurʾānic Studies, 6/1
(2003), p. 143-5 (mentioning Dutton’s paper, “Three Possibly pre-ʿUthmānic Folios of the
Qurʾān”); Fedeli, “Early Evidences”, p. 293-316.
31
At one point Alba Fedeli refers to “the so-called text of ʿUt ̠mān, which has now become
‘the standard text’ since 1924” (“Early Evidences”, p. 306). Since ʿUt ̠mān’s text became official
1268 years before that date, she has something else in mind, namely the Egyptian edition. On
another page (p. 315), she says that it was in the fourth/tenth century that the ʿUt ̠mānic text
became official; but this is wrong. Here, the author appears to conflate Ibn Muǧāhid’s canon-
ization of particular ʿUt ̠mānic readings (against other equally ʿUt ̠mānic readings alongside non-
ʿUt ̠mānic ones) with the canonization of the ʿUt ̠mānic codex (against non-ʿUt ̠mānic ones). For
some other issues, see next footnote.
32
Fedeli thinks that the phrase ʿan dīnihi in Kor 2, 217 is missing from the lower text, and
this becomes the basis for her theory that the term was added deliberately to the ʿUt ̠mānic ver-
sion. That is incorrect. First, she has missed the unmistakable lowermost part of the nūn of ʿan
which has survived the damage to the parchment. So,ʿan dīnihi was part of the text after all.
(ʿan dīnihi is present also in the Bonhams 2000 folio in Kor 5, 54.) Second, the entire sentence
wa-man yartadid minkum ʿan dīnihi fa-yamut wa-huwa kāfirun is generally illegible due to dam-
age to the parchment. It is, therefore, not clear how she is able to conclude that the words ʿan
dīnihi in this verse are missing. Presumably the fact that the lower-modifier wrote ʿan dīnihi
leads her to think that the words were not there originally; but, as shown above in footnote 12,
the lower modifier sometimes wrote words that filled the gaps created by irremediable erasure.
It is thus entirely possible that ʿan dīnihi was part of the lower text, was erased irremediably, and
then was written again by the lower modifier. Third, even if the term were missing here, there
would be no reason for considering the ʿUt ̠mānic wording as the later one, as opposed to the
earlier one. Her choice in this regard and her assumption of deliberate change betray, perhaps,
a slight measure of conspiracy-mindedness. Fourth, the scenario Fedeli constructs to explain
what may have motivated the addition of ʿan dīnihi is unclear as it stands (see “Early Evi-
dences”, p. 314).

II. The ʿUt ̠mānic Textual Tradition and the Upper Writing of Ṣanʿāʾ 1
What We Know about the Codex of ʿUt̠mān
A large number of Qurʾān manuscripts from the first and early second cen-
tury AH are extant. These are dispersed around the world in libraries, muse-
ums, mosques, and private hands. With the exception of the lower writing of
Ṣanʿāʾ 1, they all belong to the standard textual tradition, called ʿUt ̠mānic in
accordance with the traditional account of its origin, which traces it to the
recension promulgated by ʿUt ̠mān b. ʿAffān, the Companion of the Prophet
Muḥammad who ruled the Muslim empire as the third caliph during
23-35/644-56. Sometime around AD 650, ʿUt ̠mān is said to have sent copies
of the Qurʾān to Kūfa, Baṣra, and Syria and to have kept a copy with him in
Medina.33
How much of the traditional account of ʿUt ̠mān’s dissemination of the
standard Qurʾān is true? The reports that accumulated in the literary sources
about the early history of the text are not without problems. They contain
apparent contradictions and gaps in certain aspects of what transpired before
ʿUt ̠mān disseminated the text. However, there are specific elements relating
to the act of dissemination itself that can be verified. We know at least two
things. First, there is no doubt that it was ʿUt ̠mān who established the stan-
dard by sending out master copies to different cities. Second, there is strong
reason to believe that the text he established was what the sources say it was
as far as the skeletal-morphemic aspect is concerned (i.e. not including differ-
ences in spelling conventions).
Our knowledge of these fundamentals does not depend on uncritical
acceptance of reports of individuals (āt̠ār). For the first piece of our knowl-
edge, that it was ʿUt ̠mān who established the standard version, one may note
two arguments, one inspired by a passing observation of Hossein Modarressi
about the collective memory of the early communities, and the other inspired
by recent studies of Michael Cook and others that have shown the reliability
of the qirāʾāt literature. For the second proposition, that the skeletal-morphe-
mic ʿUt ̠mānic text was what the sources say it was, I refer to Cook’s findings.
To these arguments may be added the circumstantial corroboration to be
offered by the present essay: textual analysis indicates that the ʿUt ̠mānic text
preserves the Prophetic prototype better than C-1, and yet the lower writing
is older than AD 656, probably belonging to the fifteen years after the
33
For the date of this event see above, p. 344, as well as footnote 2.

Prophet’s death. Therefore, the ʿUt ̠mānic wording is older than that of another
codex already from the era of the Companions.34
How can one know that it was ʿUt ̠mān who sent the regional codices? The
first argument for this concerns collective memories. In the late first century
and early second century AH, communities remembered having received the
standard text from ʿUt ̠mān.35 It was not only the different cities that had the
same recollection, but also the different and at times clashing religious com-
munities, including the proto-Šīʿīs and the Ḫ āriǧīs. No one traced the stan-
dard version to a source other than ʿUt ̠mān. The ʿAlīds did not ascribe it to
ʿAlī. Even those who transmitted the variants found in non-standard codices
(such as the codex of Ibn Masʿūd) did not dispute the ʿUt ̠mānic origin
of the standard version.36 Reports of individuals were inconsistent on the
34
Note that my argument about ʿUt ̠mān does not rely on reports of individuals. For the
purpose of this essay, I generally reserve judgment on the accuracy of reports of individuals
unless a report receives corroboration from other types of evidence. Nonetheless, the number of
traditions that bear on the origins of the ʿUt ̠mānic recension directly or indirectly is very large,
and each report deserves detailed analysis. For a study of just one tradition, see Harald Motzki,
“The Collection of the Qurʾān: A Reconsideration of Western Views in Light of Recent Meth-
odological Developments”, Der Islam, 78 (2001), p. 1-34.
35
My argument from the collective memory of early communities is inspired by Hossein
Modarressi, “Early Debates on the Integrity of the Qurʾān: A Brief Survey”, Studia Islamica, 77
(1993), p. 13-4. A variant of the argument was subsequently used by Fred Donner, although
not specifically with respect to ʿUt ̠mān’s standardizaton of the Qurʾān (Fred Donner, Narratives
of Islamic Origins, Princeton, NJ, Darwin Press, 1998, p. 26-8); cf. Hugh Kennedy, The Prophet
and the Age of the Caliphates, New York, Longman, 1986, p. 357.
36
A report in Ibn Šabba l-Numayrī, Taʾrīḫ al-Madīna, Qum, Dār al-fikr, I, p. 7-8, states that
the Umayyad governor of Kūfa, al-Ḥ aǧgā̌ ǧ (d. 95/714), had a codex copied and sent to the main
cities. The report adds that, later, the ʿAbbāsid caliph al-Mahdī also sent a codex to Medina. The
report explicitly acknowledges the prior codex of ʿUt ̠mān. Al-Ḥ aǧgā̌ ǧ’s codex belonged to the
ʿUt ̠mānic textual tradition. This is indicated by a Baṣran report listing the eleven alleged “changes”
al-Ḥ aǧgā̌ ǧ made to the Qurʾān. Nine of the eleven variants involve just single characters (Ibn Abī
Dāwūd, al-Maṣāḥ if, p. 49-50). The differences are well within the range of variations one expects
to emerge naturally within a textual tradition. A close study of the variants shows that the Baṣran
author of the report had simply assumed that one particular Baṣran copy belonging to the stan-
dard text type represented the original text sent out by ʿUt ̠mān. He thus naively assumed that the
eleven differences with al-Ḥ aǧgā̌ ǧ’s codex represented changes made by the despised governor. In
any case, there is no chance that al-Ḥ aǧgā̌ ǧ could have dislodged the various regional branches
of the ʿUt ̠mānic textual tradition especially outside Iraq, and there is no evidence that he
attempted to do so either in or outside Iraq; but there is evidence that if he did try, he failed.
Muslim reports about his use of coercion concern Ibn Masʿūd’s text type. Furthermore, nobody
ever confused his codex with the original copy of ʿUt ̠mān. One should also dismiss a minority
Šīʿī claim that God’s explicit references to ʿAlī were removed from the Qurʾān, as well as a similar
early ʿAbbāsid claim (reported by ʿAbd al-Masīḥ al-Kindī) about God’s explicit references to the
ʿAbbāsid and Umayyad dynasties. I intend to discuss all of the above matters in detail in a sepa-
rate article devoted to the role of al-Ḥ aǧgā̌ ǧ.

history of the text before ʿUt ̠mān’s promulgation of the standard version, but
no one doubted that it was ʿUt ̠mān who disseminated it.
It is not possible to envision an Umayyad caliph establishing the standard ver-
sion. To assume so would require explaining not only how the Umayyads man-
aged to erase their empire-wide intervention from the memories of their subjects,
whose statements have reached us in large numbers, in various cities, tribes, and
sects, but also how it was possible to induce the same false memory about
ʿUt ̠mān among all, including the dissidents who preferred ʿAlī to ʿUt ̠mān.
To be sure, entire populations can believe in falsehoods. But there are facts
of a public nature that one expects people to know. For example, populations
know at any given time who their king or caliph is. The standardization of
the Qurʾān is something that people in the main cities would have known
about. Imagine a city receiving a codex of the Qurʾān from the caliph, one
intended to replace the codices in the hands of its dwellers. This would have
been a very public event, an occurrence of which the whole society would
have been aware, a sensitive affair that people would have talked about. The
city would know under which caliph it got its Qurʾān. Now give free rein to
imagination: suppose it was the caliph ʿAbd al-Malik who standardized the
Qurʾān. The question is: how was he able to persuade everybody that they
did not receive the codex when they thought they got it? How could he con-
vince even the communities teeming with rebels, or those who did not care
much for ʿUt ̠mān?
If we suppose that the recollection of the different cities, sects, and tribes
changed over time, replacing the name of ʿAbd al-Malik with ʿUt ̠mān, then it
becomes impossible to explain why the memories of all cities, sects, tribes,
and individuals changed in exactly the same way. Relatively decentralized, and
continually facing rebellion and dissension, the Umayyads were hard-pressed
to preserve their political authority over their domain. Damascus did not
attempt to micromanage religious opinion across the empire, nor would it
have succeeded in creating a consensus to replace an earlier consensus if it had
tried to. Even in the much more centralized times that were to follow under
the ʿAbbāsid caliph al-Maʾmūn, the attempt to fix theological dogma would
prove futile. Revisionist back-projection is a common thing in both pre-mod-
ern and contemporary understandings of history. But such a back-projection
would have failed to change without any trace of dissent what would have been
common knowledge in several cities that were home to numerous clashing or
rebellious communities. This observation becomes even more salient when
one notes that on the vast majority of legal questions, each city had a distinct
profile. The routine disagreements among the cities shows a large measure of
independent religious development during the first century of Islam.

While the above argument concerns collective recollections of a momen-

tous occurrence, my second argument is about individuals’ memories of
minutiae, memories that recently have been shown to be accurate. The com-
munities had more than just a recollection of the source of their Qurʾāns.
There were established scholarly traditions of reading them that varied from
scholar to scholar and from region to region. Scholars began to document
these traditions by compiling and classifying the differences in pronunciation
and orthography, and variations in the way sūras were divided into verses
(viz. the locations of the verse separators). Their efforts culminated in the
genre of qirāʾāt, the so-called “Readings” literature. The extant books in this
field were written much later than the periods they describe, relying on
reports that purportedly had been handed down for several generations. Yet
recent research, to be discussed below, has shown that when it comes to the
minutiae of textual variations, the literature carries substantially accurate
information about the first- and early second-century phase. (To be sure,
none of this implies that the genre is immune to inaccuracies and errors of
transmission or that it captured everything about the early period.)
If the qirāʾāt literature is substantially credible with regard to minutiae,
that is all the more reason to accept the historicity of the broad framework it
takes for granted. To give a whimsical yet apt analogy, imagine a person who
alleges that there is an eight-hundred-pound gorilla in a room. It is unreason-
able to say that his claims are substantially right about fine features of the
primate—the color of its fur, the sizes of its toes, etc.—but wrong about its
very existence. A dogmatic skeptic could devise a scenario where this could
happen, but it would not be the natural conclusion. By the same token, the
most natural way to explain why the qirāʾāt genre is accurate in many of its
particulars is to accept the truth of its core assumption that the standard tex-
tual tradition goes back to ʿUt ̠mān. For instance, it is hard to imagine that
people could remember accurately the archetypal regional codices at the level
of individual words, as Cook has shown, but get wrong the name of the caliph
who disseminated those codices. It would be as if tradition preserved accu-
rately all the minutiae of the Battle of the Bulge from WWII but projected
the event back to WWI.
Among recent studies that give credibility to the qirāʾāt genre, two are partic-
ularly important for the purposes of my study: those of Cook and Yasin Dutton.
Cook analyzed the thirty-six reported differences in the codices ʿUt ̠mān sent to
the cities—the original codices, as opposed to their less pristine offspring.37
37
Michael Cook, “The Stemma of the Regional Codices of the Koran”, Graeco-Arabica, 9-10
(2004), p. 89-104.

These variants must have been introduced in Medina as the texts were copied
off one another before being dispatched to the cities. They consist of small
changes to the skeletal text that actually make a difference in pronunciation,
usually changing one word to another (hence my label, “skeletal-morphemic”
changes), as opposed to changing merely the spelling of a word. Typically,
has wa-qāla ‫“( ﻭﻗﺎﻝ‬and he said”) whereas others have qāla ‫“( ﻗﺎﻝ‬he said”). In
the meaning does not change. For example, in Kor 7, 75 the codex of Syria
a few cases, the meaning is affected slightly. For example, the codex of Syria
has wa-yanšurukum ‫“( ﻭﻳﻨﺸﺮﻛﻢ‬makes you spread”) in Kor 10, 22 whereas oth-
ers have the graphically similar term wa-yusayyirukum ‫“( ﻭﻳﺴﻴﺮﻛﻢ‬makes you
journey”).
Cook’s investigation of the list of skeletal-morphemic variations which,
according to early scholars, distinguished the ʿUt ̠mānic codices, allowed him
to reconstruct the genetic relationships between and among the codices. The
variations among the cities follow strikingly regular patterns. For example,
the codex of Medina has no variant that is unique to it. Some variants belong
uniquely to it and Syria, and others uniquely to it and Baṣra. But it never
agrees with Kūfa against Syria or Baṣra.38 When such patterns are taken into
consideration, the upshot is that of the well over 100 conceivable stemmata
that could describe which codex was copied from which, only several closely-
related ones fit the data, and among these some are more probable than
others.39 On the one hand, the great regularity of the patterns requires an
explanation other than chance, and, on the other hand, it is difficult to
explain why the texts fit this very specific set of stemmata if not as a result of
having emerged in accordance with one of them. Cook argues convincingly
against the possibility that the descriptions of these codices were forged by
pointing out that the variants of four cities show no sign of cross-contamina-
tion. To create the appearance of non-contamination would have required
skills that the scholars did not have, namely knowledge of the logic of stem-
mata.40 From this, Cook infers that “we have to do with genuine transmis-
sions from an archetype”,41 thus showing the correctness of the reports about
the regional codices. The historicity of the received data, in turn, is “a testi-
mony to the continuing accuracy of the transmission of the variants”42 in the
qirāʾāt literature. (Elsewhere, he reaches a similarly positive conclusion about
38
Ibid., p. 92-3.
39
Ibid., p. 103.
40
Ibid., p. 103-4.
41
Ibid., p. 104.
42
Ibid., p. 103.

the accuracy of the literary sources in describing the personal codex of a con-
temporary of ʿUt ̠mān.)43
Whereas Cook’s corroboration of the qirāʾāt variants turns on the analysis
of their patterns of agreement and disagreement, Dutton and a number of
other scholars compare the literature with external evidence in the form of
manuscripts.44 For example, the variants of the ʿUt ̠mānic Qurʾān in the Brit-
ish Library, which is among the earliest set of manuscripts, point precisely
and redundantly to the Reading tradition ascribed to Ḥ imṣ, a town located
north of Damascus.45 The match between the text of the ancient folios and
the reported attributes of the Ḥ imṣī reading in the qirāʾāt genre is improbable
a priori in the way it is improbable for two randomly picked pieces of a jig-
saw puzzle to fit. The match provides the kind of mutual corroboration akin
to that of the conjoined pieces of the puzzle, in that each piece sheds light on
the other: the manuscript tells us that the reported Ḥ imṣī reading found in
the qirāʾāt literature preserves genuine information about the earliest class of
manuscripts, and the qirāʾāt tell us that the folios were the work of a scribe
following a textual tradition associated with the town of Ḥ imṣ. Likewise, as
shown by Tayyar Altıkulaç, et. al., a nearly complete early codex in Istanbul
reveals a match with the qirāʾāt-reported skeletal-morphemic patterns of
Medina, while the skeletal-morphemic text of the well-known Tashkent
43
Michael Cook and M.M. al-Aʿẓamī discuss a copy of the Qurʾān belonging to the Medi-
nese Mālik b. Abī ʿĀmir al-Aṣbaḥī (d. 74/693). His grandson, the famous jurist Mālik b. Anas
(94-179/712-795) showed it to his students, who recognized that its text matched the muṣḥ af
of Medina up to sūra 40 and then switched to the muṣḥ af of Baṣra. This Qurʾān was decorated
with silver, possessed ornamental bands in black ink serving as sūra dividers, and separated adja-
cent verses with a dot. According to the grandson, this Qurʾān was copied when ʿUt ̠mān had
the codices written. Attempting to explain the composite nature of the codex, Cook considers
the possibility that it was damaged at some point and its last fifth was recopied from a codex
that had reached Medina from somewhere else. This scenario is more complex than it needs to
be. Mālik b. Abī ʿĀmir was a scribe in ʿUt ̠mān’s project of standardization. He thus had access
to the muṣḥ afs before they were sent out to the cities. If he made his personal copy before the
muṣḥ afs were dispatched, he would have had to go where these muṣḥ afs were kept. There is no
reason for expecting that in each sitting he picked up the same muṣḥ af to copy. See Michael
Cook, “A Koranic Codex Inherited by Mālik from his Grandfather”, in Proceedings of the Sixth
International Congress on Graeco-Oriental and African Studies, Graeco-Arabica, ed. Vassilios
Christides and Theodore Papadopoullos, VII-VIII, Nicosia, 7-8 (1999-2000), p. 93-105;
al-Aʿẓamī, The History of the Qurʾānic Text, p. 97-8, 170-2.
44
Rabb, “Non-Canonical Readings”, p. 84-127; Yasin Dutton, “An Early Muṣḥ af According
to the Reading of Ibn ʿĀmir”, Journal of Qurʾānic Studies, 3/1 (2001), p. 71-89; id., “Some
Notes on the British Library’s ‘Oldest Qurʾān Manuscript’ (Or. 2165)”, Journal of Qurʾānic
Studies, 6 (2004), p. 43-71; id., “An Umayyad Fragment of the Qurʾān and its Dating”, Journal
of Qurʾānic Studies, 9/2 (2007), p. 57-87.
45
See Dutton, “Some Notes”; Rabb, “Non-Canonical Readings”. Both are cited in the last
footnote. Rabb introduces some corrections to Dutton’s work.

Qurʾān matches that of Kūfa.46 Muḥammad Muṣt ̣afā l-Aʿẓamī has conducted
similar comparisons in the latest edition of his book on the Qurʾān.47 How-
ever, not every Qurʾān matches the data of qirāʾāt. Naturally, the manuscripts
can accumulate noise (i.e. changes due to a copyist’s discretion, e.g. his choice
of spelling conventions, or error) and contamination (influence from another
branch of the tradition); and these could accumulate in the course of hun-
dreds of copyings. Verse divisions are particularly susceptible to such changes.
For example, a single Ḥ iǧāzī folio whose image was published recently dis-
plays numerous deviations in the way the verses were divided when compared
with the data in the qirāʾāt literature.48 Likewise, Déroche has noted that MS
Arabe 328 in the Bibliothèque nationale de France (BNF) does not match
neatly any of the qirāʾāt-reported patterns.49
To help locate the upper text of Ṣanʿāʾ 1, the present essay follows in the
footsteps of Dutton. I find a fit with information in the qirāʾāt literature. It
will emerge that the likelihood that this match is due to chance is nil.
The Date of the Upper Writing of Ṣanʿāʾ 1
Why was the lower script erased and written over? Perhaps the original owner
deemed the lower writing obsolete after ʿUt ̠mān’s standardization. This would
be likely if the codex belonged to a mosque. Perhaps after the original owner
died, the new owner wished to replace the text with the ʿUt ̠mānic version.
Alternatively, it is possible that the original codex needed to be written over
because it was damaged or worn out. This would fit the fact that a narrow
horizontal strip along the bottom edge of some of the leaves was cut and dis-
carded before the text was written over. The question remains open.
When was the upper layer written? Art-historical and paleographic consid-
erations assign the upper writing to the earliest stratum of extant Qurʾān
manuscripts; but, at present paleographic knowledge is not refined enough to
allow dating the manuscripts in this earliest group relative to one another.
Such methods establish only that the upper writing could be from the first
half of the seventh century AD, the second half of that century, or (though
46
Tayyar Altıkulaç and Ekmeleddin İhsanoğlu, Hz. Osman’a izâfe edilen mushaf-ı şerif:
Topkapı Sarayı Müzesi Nüshası, Istanbul, İslam Târih, Sanat ve Kültür Araştırma Merkezi,
1428/2007. I owe this reference to Shahzad Bashir.
47
Al-Aʿẓamī, The History of the Qurʾānic Text, p. 182-8.
48
David Roxburgh, Writing the Word of God: Calligraphy and the Qur’an, The Museum of
Fine Arts, Houston, 2007, p. 6-7.
49
Déroche, La transmission écrite du Coran, p. 77-102. Déroche notes that the verse division
patterns of the BNF manuscript are slightly closer to the Ḥ imṣī tradition reported in the
sources.

perhaps less likely) from early eighth century.50 These methods thus leave any
date between AD 632 and AD 700 feasible. Some might conjecture that the
upper layer was written shortly after ʿUt ̠mān’s standardization around AD
650. The fact that all the other known early manuscripts are ʿUt ̠mānic sug-
gests that the standard Qurʾān became predominant quickly. So, there may
have been a tendency to replace non-ʿUt ̠mānic Qurʾāns. That said, literary
evidence makes it clear that not everybody gave up non-ʿUt ̠mānic codices,
that non-ʿUt ̠mānic variants were handed down by prominent scholars, and
that physical copies of such codices were extant in the early centuries.51
Finally, one may ask where the upper layer was written. To help answer that
question, I set out to locate the upper writing within the “reading” traditions
of the cities as described in the qirāʾāt literature. To do so, one may use two
resources: the orthographic variants of the cities and their varying systems for
dividing the passages into verses.
The Provenance of an ʿUt̠mānic Qurʾān: Results and Methods
Recent attempts by Yasin Dutton and others to bring the data of the qirāʾāt
to bear on several of the earliest available ʿUt ̠mānic manuscripts have proved
a great success. In this essay, I follow in their methodological footsteps to
shed light on the provenance of the ʿUt ̠mānic upper writing of Ṣanʿāʾ 1.52 The
method is to compare the data of the manuscript with the limited number of
patterns reported in the qirāʾāt literature for different cities and see if there is
a match. (The qirāʾāt genre was defined above, on p. 367.) It emerges that the
upper text could not have been written in Damascus, Ḥ imṣ, Kūfa, or Baṣra,
but is consistent with the traditions of Mecca or Medina. It was probably
written in one of those cities, in Yemen, or in Egypt.
50
On paleography and art history, see these references and the sources cited therein: Blair,
Islamic Calligraphy, p. 101-40; Déroche, “New Evidence about Umayyad Bookhands”, p. 611-
42; Ṣalāḥ al-Dīn al-Munaǧǧid, Dirāsāt fī taʾrīḫ al-ḫaṭt ̣ al-ʿarabī mund̠u bidāyatihi ilā nihāyat
al-ʿaṣr al-umawī, Beirut, Dār al-kitāb al-ǧadīd, 1972.
51
See e.g. Ibn Abī Dāwūd, al-Maṣāḥ if, p. 14-8; Ibn al-Nadīm, Abū l-Faraǧ Muḥammad b.
Abī Yaʿqūb Isḥāq, Kitāb al-Fihrist, ed. Riḍā Taǧaddud, n.d., p. 29; Rāmyār, Tārīḫ-i Qurʾān,
p. 362-4, 382-4, 389-92.
52
The contributions of Yasin Dutton and Intisar Rabb are cited in footnote 44 above (Rabb
refines the work of Dutton). For related approaches by Tayyar Altıkulaç, Muḥammad Muṣt ̣afā
l-Aʿẓamī, and now François Déroche see footnotes 46, 47 and 49. For the broad significance of
such studies, see the discussion above beginning on p. 369. To my knowledge, the earliest schol-
ars known to have used this method are those from the second AH/eighth century AD who
applied it to a manuscript belonging to the grandfather of Mālik b. Anas, for which see the sec-
ondary sources cited in footnote 43.

To appreciate just how formidable this approach can be, let us consider a
simplified hypothetical example. Suppose one is interested only in differences
in the way passages are divided up into verses, and suppose further that in a
partial Qurʾān manuscript, there are twenty points in the text at which the
qirāʾāt literature reports disagreements over the existence of a verse separator.
The total number of possible ways in which a scribe may assign verse separa-
tors to these twenty points is about a million, specifically 1,048,576 (i.e. 220).
Yet the qirāʾāt literature offers only seven patterns, two associated with Med-
ina, and five with Mecca, Kūfa, Baṣra, Damascus, and Ḥ imṣ. That means that
the a priori probability that any manuscript fits one of these seven patterns at
those twenty spots is virtually nil (about 0.000007). Therefore, if a manu-
script does fit one of these profiles, the match is so striking that it cannot be
explained by chance.
It should be noted that the qirāʾāt were not intended to describe everything
found in manuscripts, including errors; rather, they aimed to convey the
knowledge of experts about what the correct text is likely to be, knowledge
that was only partly based on analyses of contemporaneous manuscripts. This
makes their affinity with some of the earliest available manuscripts notewor-
thy. One would not necessarily have expected a close correspondence. After
all, in the earliest phase, many a copyist’s skill set need not have exceeded lit-
eracy and the possession of good handwriting, whereas the bar for being a
scholar would have been set higher.53 Even the most meticulous scribe’s text
could be only as good as the manuscript he was copying, which would have
incorporated the changes accumulated in the course of the hundreds of ear-
lier copyings of careful and careless scribes.
With regard to verse divisions, in particular, a significant amount of accu-
mulated noise (copyists’ errors and exercises of discretion) eventually led to a
very complex reality of which the qirāʾāt capture but a cross-section. So, one
cannot take it for granted that there will be a match—and indeed some man-
uscripts do not exhibit a close match when it comes to patterns of verse divi-
53
There is a substantial amount of evidence for mistakes of individual copyists in Qurʾānic
manuscripts (nonsense-generating haplography, dittography, verse-separator placement, spell-
ing mistakes, etc., and singular readings). This, of course, does not mean that every scribe was
careless. There is a parallelism with the New Testament textual tradition: there were plenty of
sloppy copyists in the early times, although “the case of P75 shows that at least some scribes were
capable of care” (J.R. Royse, “Scribal Tendencies in the Transmission of the Text”, in The Text of
the New Testament in Contemporary Research: Essays on the Status Quaestionis, ed. B.D. Ehrman
and M.W. Holmes, Grand Rapids, 1995, p. 239-52). Over time standards improved; thus,
“early scribes did not exercise the care evidenced in later transcriptions” (ibid.). For two
instances of scribal incompetence so egregious as to be entertaining, see Bruce M. Metzger, The
Text of the New Testament: Its Transmission, Corruption, and Restoration, 3rd edition, Oxford,
1992, p. 194-5.

sion. And yet, remarkably, there are instances of a close correspondence

between manuscripts and the qirāʾāt. In such cases, the qirāʾāt can help reveal
the provenance of manuscripts.
How to Compare Manuscripts and the Qirāʾāt (“Readings”) Literature
What kinds of variants can be found in manuscripts, and which ones are
treated in the qirāʾāt genre? ʿUt ̠mānic Qurʾān manuscripts could encompass
several different types of variants: differences in vowel markings and hamza
markings in those manuscripts that do have such markings, variations in the
etal shape (such as the letter ‫ ﺏ‬versus ‫( )ﺕ‬cf. first row of Table 5), spelling
small dashes distinguishing one consonant from another one of a similar skel-
differences that do not change a word or its pronunciation, spelling differ-
(e.g. ‫ ﺍﻭﻻ ﺍﻻﻟﺒﺐ‬vs. ‫)ﺍﻭﻟﻰ ﺍﻻﻟﺒﺐ‬, and the use of different words that have similar
ences that do change the pronunciation of a word but not its meaning
shapes. Among these types, some (such as those involving diacritics and
vowel markings) can be recognized less frequently or not at all in the earliest
manuscripts. For example, the upper writing of Ṣanʿāʾ 1 does not have vowel
markings.
The qirāʾāt literature provides systematic information on all the above types
of variants, but it is not as thorough for graphical (or skeletal, rasm) variants
that neither potentially change one word into another nor make a difference
in its pronunciation (second row of Table 5). In the English language, such
variations would be akin to the difference between the spellings “adviser” and
“advisor”, or between “behavior” and behaviour”. Though understandably
relatively marginal in the qirāʾāt literature, variants of this type constitute the
largest category of variations in the manuscript evidence. These variations
owe, most of all, to the lack of consistency in the way the long vowels were
represented in early Arabic writing, as different copyists used different spell-
ing practices, and sometimes the same scribe switched from one spelling of a
word to another.54 Among such orthographic variants, by far most involve
sented by the letter alif (‫ )ﺍ‬or the letter yāʾ (‫)ﻯ‬, yielding either ‫ ﻋﻼ‬or ‫ ﻋﻠﻰ‬as
the long vowel ā. For example, the ā sound in the word ʿalā could be repre-
alternative spellings. More often, however, the choice was between denoting
the ā sound by the letter alif and simply not representing it. Thus the mor-
pheme kātib (in Kor 2, 283) could be written as k - ā - t - b (‫ )ﻛﺎﺗﺐ‬or k - t - b
(‫)ﻛﺘﺐ‬.
54
E.g. al-Aʿẓamī, The History of the Qurʾānic Text, p. 145; for examples from early non-
Qurʾānic manuscripts, see Ibid., p. 163-4.

Table 5. The utility of the literary sources for determining the provenance
of a manuscript fragment in the ʿUt ̠mānic textual tradition depends on
the types of variants found in the fragment.
Type of variant Presence in Information from the Usefulness of
Manuscripts Literary Sources Qirāʾāt for Assigning
Provenance
1. Different Earliest Extensive Not useful for

vowels or different manuscripts manuscripts that lack
consonant- have no or a diacritics or vowel
distinguishing limited number marks at spots of
diacritics of diacritics and interest
vowel marks
2. Skeletal Very frequent Limited Not very useful
differences
that do not
necessarily change
pronunciation
3. Skeletal Frequent to a Precise & accurate for Useful if some of the
differences that degree consistent the original codices 36 spots where the
necessarily change with written sent out by ʿUt ̠mān original ʿUt ̠mānic
pronunciation transmission (Mecca, Medina, codices disagree
(“skeletal- Kūfa, Baṣra, Syria). are included in the
morphemic”) Lacking vis-à-vis manuscript fragment
changes accumulated
later.
4. Differences in Frequent Lacking for the Useful if the verse
the locations of original codices sent divisions in the
verse divisions out by ʿUt ̠mān. Some manuscript match
of the later branches one of the systems
of the tradition are recorded for the
recorded (Mecca, systems. (There is not
Medina, Damascus, always a match.)
Ḥ imṣ, Kūfa, Baṣra).
Not all later schemes
are recorded.

Note that in the second case (k-t-b), with the alif absent, without the linguis-
tic context the reader would not know whether to read the term as kataba,
kutiba, kutub, kitāb, kātib, or kuttāb, all seemingly valid choices; but here
(Kor 2, 283) the context removes ambiguity. In the case of the first spelling
(k-ā-t-b), by contrast, the orthography itself fixes the first vowel as ā. There
are other cases where even the context does not completely disambiguate a
word. In these instances, the qirāʾāt literature gives the variations in pronun-
ciation that may have existed among the early authorities.
What types of variation can help one check manuscripts against the qirāʾāt
“readings”? The type of variants about which the qirāʾāt have the most to say,
namely those concerning different vowellings of the skeletal text, is of the
least use for unvowelled manuscripts. Likewise, the largest category of vari-
ants found in manuscripts, namely spelling variations that do not necessarily
change a word (e.g. k-ā-t-b vs. k-t-b for kātib in Kor 2, 282), is not very use-
ful, because the qirāʾāt literature does not give a thorough description of
regional spelling practices. In any case, it is highly likely, given that spelling
conventions were not standardized, that different spellings of a word did not
time and place may spell the word ʿalā differently (‫ ﻋﻠﻰ‬vs. ‫)ﻋﻼ‬, as did the two
usually signify regional variation. For example, two scribes from the same
copyists of the upper layer of Ṣanʿāʾ 1. (Scholars did, however, record over
200 instances in which codices of the main cities agreed on the spelling of a
word.55 They also recorded some cases in which the codices of one city
seemed to agree.56 This suggests that further research might identify certain
words of which the spelling could have diagnostic value.)
This leaves mainly two types of variants: (1) skeletal-morphemic differ-
ences, (2) differences in the way verses are demarcated. I take these up in turn.
Skeletal-Morphemic Variants in the ʿUt̠mānic Textual Tradition
How should one go about comparing the skeletal text (rasm) of a manuscript
with the qirāʾāt, and what can one learn from such a comparison? This sec-
tion considers the relatively small subset of such variants that necessarily
change a word, even if only slightly, or make a difference in the way it is pro-
nounced. It is convenient to have a label for them; and so I call them “skele-
tal-morphemic variants”. In other words, the present discussion is limited to
the skeletal written form, and excludes orthographic variations that do not
55
For Ibn Abī Dāwūd al-Siǧistānī’s list of words of which the spelling was agreed upon, see
his Kitāb al-Maṣāḥ if, p. 105-17. As could be expected, relatively few of these involve the letter
alif. This confirms that the use of alif was inconsistent.
56
E.g., for Medina, Ibid., p. 104.

necessarily change a word or its pronunciation (third row of Table 5). The use
of fa- instead of wa- is an example of a skeletal-morphemic variation; but dif-
ferences in the spelling of a word—e.g. the spelling k - t - b (‫ )ﻛﺘﺐ‬instead of
k - ā - t - b (‫ )ﻛﺎﺗﺐ‬for kātib in Kor 2, 283—are not.
As discussed above (p. 368), Cook has shown that the qirāʾāt literature
provides authentic and accurate information about differences of this kind
among the master codices that ʿUt ̠mān sent to the cities.57 If the ʿUt ̠mānic
textual tradition is one that evolved over time, we have the splendid fortune
of being able to say that we already know the initial state in this chain of suc-
cession; we know the roots and the trunk in the genetic tree diagram, the
original ʿUt ̠mānic master codices. This holds, though, only at the skeletal-
morphemic level, not always for the less important cases where an ortho-
graphical difference does not necessarily make a difference in pronunciation.
For example, this often does not hold for the spellings of words. In many
cases, we do not know the original spellings of words in the codices ʿUt ̠mān
sent out. Most manuscripts, on the other hand, constitute evidence of the
later developments of the written text: namely, the leaves and branches of the
ʿUt ̠mānic textual tradition. For once, then, scholars need not toil to recon-
struct a prototype from the later realizations. Thanks to the efforts of the ear-
liest Qurʾān specialists, one is not left at the mercy of the copyists and the
changes they introduced in the texts.
These insights should make it clear how to approach an ʿUt ̠mānic manu-
script. If its orthography matches one of the profiles reported for Mecca,
Medina, Kūfa, Baṣra, or Syria, then the manuscript belongs to that city or to
regions following the tradition of that city. On the other hand, words incon-
sistent with the regional master codices sent by ʿUt ̠mān can be safely regarded
as later accretions. Such deviations, of course, can be handed down in a tex-
tual tradition, and therefore may have value for determining the provenance
57
Michael Cook, “The Stemma of the Regional Codices of the Koran”, p. 89-104. Where
did the knowledge of the skeletal-morphemic differences of the regional ʿUt ̠mānic archetypes
come from? Investigators may have gleaned the differences between regional codices by examin-
ing the original master copies of ʿUt ̠mān. Al-Aʿẓamī makes a case for this scenario (The History
of the Qurʾānic Text, p. 167). The details of this early scholarly activity merit further inquiry. As
another possibility, one ought to investigate whether the reported orthographical differences
were derived from the qirāʾāt, the actual scholarly regional traditions of reciting the Qurʾān.
Orthographical variants that made a difference to pronunciation would have been implicit in
the qirāʾāt (cf. Muḥammad Sālim Muḥaysin, al-Fatḥ al-rabbānī fī ʿalāqat al-qirāʾāt bi-l-rasm
al-ʿut̠mānī, Saudi Arabia, Ǧ āmiʿat al-Imām Muḥammad b. Saʿūd al-islāmiyya, 1415/1994). As a
third possibility, scholars may have conducted rudimentary textual analysis of high-quality first-
century manuscripts from different cities. Whichever method was used, Cook’s study shows
that it yielded accurate results.

of a manuscript. This is because one may more effectively assign a manuscript

to a place if one has regional profiles based on all textual characteristics,
including accumulated errors. However, the current state of scholarship is not
sufficiently advanced to allow one to put accumulated errors to such a use.
Verse Divisions in the ʿUt̠mānic Textual Tradition
By verse-division variants, I mean differences in the ways in which different

cities or individuals divided the sūras into verses. What I present here is what
the literary sources (qirāʾāt literature) have to say about the variations in such
divisions, not necessarily what one always observes in manuscripts—manu-
scripts can possess divisions not described in the qirāʾāt literature or lack verse
divisions reported therein, and such deviations may be few or many. One
may compare the data of a manuscript with the reported verse division
schemes to see if there is a match (last row of Table 5).
The literary sources (qirāʾāt reports) describe some of the traditions of
dividing verses in Mecca, Medina (two traditions), Kūfa, Baṣra, Damascus,
and Ḥ imṣ. These traditions may agree or disagree at a particular point. In
thirty-eight of the sūras, no such difference arose: the cities were unanimous
on how to divide the sūras into verses. Ten sūras contain a single disputed
division apiece, while most of the rest of the sūras have several disputed divi-
sions. Sūra 20 stands out with over twenty disputed cases. Interestingly, the
density of disputed points is greater in sūras with shorter verses. The similari-
ties and differences among the verse-division schemes reported in the qirāʾāt
reveal something about how they came about:
On the one hand, the reported verse-division variants are fewer than what
one would expect if different cities had divided the sūras independently.
Despite the fact that some verse-endings are “obvious” cases, it seems that if
two persons independently used discretion to insert verse divisions in the
undivided text, the resulting disparity would be great compared to the differ-
ences among the schemes reported in the qirāʾāt literature. This shows that
the verse-division traditions are linked genetically: they were derived from
one another and/or from common sources. Furthermore, divisions that are
not disputed, i.e. the vast majority of divisions, can be safely ascribed to a
prototype.
On the other hand, verse-division variants are much more numerous
than the skeletal-morphemic differences of the regional ʿUt ̠mānic archetypes.
The gap might reflect, in part, the fact that the reports on orthography
describe the original codices that ʿUt ̠mān disseminated, whereas the reports
on verse division concern the regional traditions that had evolved from those

codices. There is, in other words, a time gap allowing, in principle, for grad-
ual divergence from the original state. However, this alone does not explain
the large number of variants. The more important factor is that verse divi-
sions inherently were more susceptible to individual discretion, transmission
error, and/or cross-contamination. The scribes must have considered the
skeletal-morphemic text to be more important than the way in which it was
divided up, leading them to exercise less care in transmitting the separator
marks exactly as they found them.
The presence of accumulated “noise” (transmission error and exercise of
individual discretion) and “contamination” (the influence of one city’s textual
tradition upon another) in the verse-division systems described by the qirāʾāt
makes it relatively difficult to reconstruct the genetic relationships between
the verse-division practices of the cities. (This is in contrast to the straightfor-
ward manner in which such relationships can be detected based on ortho-
graphic variants.)58 While such a reconstruction falls outside the scope of the
present essay, two unmistakable patterns may be pointed out: First, Mecca,
Medina I, and Medina II fall into a cluster apart from Kūfa, Baṣra, and Syria.
(Syria includes Damascus and Ḥ imṣ, whose systems were relatively close to
each other; and Medina I and II refer to two different systems in Medina.)
The verse division systems of any two cities in this cluster are closer to each
other than to that of any city not in the cluster. Second, surprisingly, Medina
I is closer to Mecca than to Medina II. Similarly, Medina II is closer to Mecca
than to Medina I.59
The greater number of verse division variants (as compared to skeletal-
morphemic variants) makes them a particularly useful diagnostic tool in cases
in which only a small part of a codex has survived. This is because, on aver-
age, a passage contains more disputed verse divisions than disputed features
of the skeletal text. Such variants can be very helpful in pinning down the
provenance of a manuscript. For example, the fact that the British Library’s
“oldest Qurʾān manuscript” has been assigned to the tradition of Ḥ imṣ shows
that a high level of discrimination is possible, given the closeness between the
verse demarcation schemes of Ḥ imṣ and Damascus.60
58
Ibid.
59
There are four possible stemmata: (1) M2 ← Mec ← M1; (2) M1 ← Mec ← M2;
(3) (M1 ← Mec) and (M2 ←Mec); (4) (Mec ← M1) and (Mec ← M2). This last is unlikely,
since it does not explain why M1 and M2 belong to the same cluster. Here, M1, M2, and Mec
refer respectively to Medina I, Medina II, and Mecca.
60
See the sources cited in footnote 45.

Verse Divisions of the Upper Text of Ṣanʿāʾ 1
By checking for the presence or absence of disputed verse separators, I now

determine if the upper text matches any of the verse division traditions
reported in the literary sources.
Sūra 2, 265-27761 and 2, 277-28662

These passages do not help pin down the text. None of the eleven disputed
verse endings reported by al-Dānī or others for sūra 2 occur in the passages
found in the leaves.63
The upper text has two verse divisions not reported in al-Dānī, at wa-lā
tayammamū l-ḫabīt̠a minhu tunfiqūna (2, 267) and at wa-l-muʾminūna (2,
285). The ten-verse marker designating verse 270 appears immediately after
wa-llāhu bi-mā taʿmalūna ḫabīrun (2, 271). There is another ten-verse marker
at lā yuẓlamūna (2, 281). One could try to see how the leaf matches up with
the verse totals reported for the cities, but given the unusual length of the
sūra and the two anomalous additional verse divisions already observed,
chances are good that the manuscript would have possessed (or lacked)
another verse division anomalously, making any match misleading.
61
Stanford ’07.
62
The David folio.
63
The disputed points in sūra 2 are as follows:
alif lām mīm (2, 1): Kūfa alone counted it as a verse ending;
ʿad̠ābun alīmun (2,10): Syria alone counted it;
innamā naḥ nu muṣliḥ ūna (2, 11): Syria alone did not count it as a verse ending;
illā ḫāʾifīna (2, 114): Baṣra alone counted it;
yā ūlī l-albābi (2, 197): Medina I and Mecca alone did not count it;
mā lahu fī l-āḫirati min ḫalāqin (2, 200): Medina II alone did not count it;
mād̠ā yunfiqūna (2, 219): Medina I and Mecca alone counted it;
laʿallakum tatafakkarūna (2, 119): Medina II, Kūfa, and Syria alone counted it;
qawlan maʿrūfan (2, 235): Baṣra alone counted it;
al-ḥ ayyu l-qayyūmu (2, 255): Medina II, Baṣra, and Mecca alone counted it;
min al-ẓulumāti ilā l-nūri (2, 257): Medina I alone counted it.
The sūra has 285 verses according to Medina, Mecca, and Syria, 286 for Kūfa, and 287 for
Baṣra. See al-Dānī, al-Bayān fī ʿadd āy al-Qurʾān, Kuwait, Dār al-našr, 1414/1994, p. 88, 89, 91,
95-6, 100-1, 108, 140-2. Cf. Anton Spitaler, Die Verszählung des Koran nach islamischer Überlie-
ferung, Munich, Verlag der Bayerischen Akademie der Wissenschaften, 1935, p. 32-4.

Sūra 6, 49-6164 and 6, 149-5965

None of the four disputed verse endings reported by al-Dānī for sūra 6 occur
in the passages found in the leaves.66 Fortunately, the verse count implied by
the ten-verse markers in the upper text sheds light on the verse divisions.
There are ten-verse separators at yafsuqūna (6, 49), mubīnin (6, 59), and
aǧmaʿīna (6, 149), indicating fifty, sixty, and 150 verses respectively. These
numbers would result if there were a division at al-ẓulumāti wa-l-nūra (6, 1).
(This is the only point disputed in this part of the sūra, the next one coming
at 6, 66.) Accepting the presence of a division at al-ẓulumāti wa-l-nūra (6, 1),
one may ask how this division scheme lines up with the city profiles. The
division at al-ẓulumāti wa-l-nūra fits a Meccan or Medinan provenance and
rules out Kūfa, Baṣra, and Syria.67 This conclusion rests on the assumption
that there are no anomalous verse divisions in the manuscript prior to 6, 49.
This is not an unreasonable assumption: the fact that the three ten-verse
markers all appear at the expected positions suggests regularity.
Sūra 20, 1-4368

It is fortunate that one image covers the first forty-three verses of sūra 20,
since this sūra has the largest number of disputed verse endings in the entire
Qurʾān—over twenty of them.69 Of these the folio image covers nine. Damas-
64
Image 152255B, UNESCO CD.
65
Image 152254B, UNESCO CD.
66
The disputed points in sūra 6 are as follows:
wa-ǧaʿala l-ẓulumāti wa-l-nūra (6, 1): Medina I & II and Mecca alone counted it as a verse
ending;
qul lastu ʿalaykum bi-wakīlin (6, 66): Kūfa alone counted it;
kun fa-yakūnu (6, 73): Kūfa alone did not count it as a verse ending;
ilā sirāṭin mustaqīmin (6, 161): Kūfa alone did not count it as a verse ending;
huwa llad̠ī ḫalaqakum min ṭīnin (6, 2): a report says that the Medinans (ahl al-Madīna),
unlike the rest, counted this as a verse ending.
The total verse count is 165 for Kūfa, 166 for Baṣra and Syria, and 167 for Medina and Mecca.
See al-Dānī, al-Bayān, p. 89, 91-2, 99, 151-4. Cf. Spitaler, Verszählung, p. 36-7.
67
There is a verse separator at yaṣdifūna in 6, 157, but it looks like a scribal error. It appears
awkwardly mid-sentence. It was apparently confused with the real verse ending, which is at
another occurrence of the word yaṣdifūna. The scribe’s eye appears to have leaped to a different
occurrence of the same word, a common type of scribal error.
68
Image 152256B in UNESCO CD.
69
The disputed verse endings for sūra 20 are as follows:
ṭāhā (20, 1): Kūfa alone counted it as a verse ending;

nusabbiḥ aka kat̠īran (20, 33) and wa-nad̠kuraka kat̠īran (20, 34): Baṣra alone did not
count these two spots as verse endings;
fa-qd̠ifīhi fī l-yammi (20, 39): Ḥ imṣ alone counted it;

cus is ruled out four times—by the absence of verse separators at taqarra
ʿaynuhā wa-lā taḥ zan (20, 40), wa-fatannaka futūnan (20, 40), fī ahli Madyana
(20, 40), and wa-ṣtạ naʿtuka li-nafsī (20, 41). Ḥimṣ is ruled out four times—by
the absence of verse separators at fa-qd̠ifīhi fī l-yammi (20, 39), wa-fatannaka
futūnan (20, 40), and wa-ṣtạ naʿtuka li-nafsī (20, 41), and by the presence of a
verse separator at maḥ abbatan minnī (20, 39). Kūfa is ruled out three
times—by the absence of a verse separator at Ṭāhā (20, 1) (which can be
inferred by the location of the ten-verse marker)70 and wa-ṣtạ naʿtuka li-nafsī
(20, 41), and by the presence of a verse separator at maḥ abbatan minnī (20,
39). Baṣra is ruled out three times—by the presence of verse separators at
nad̠kuraka kat̠īran (20, 34) and maḥ abbatan minnī (20, 39), and by the
absence of a verse separator at wa-fatannaka futūnan (20, 40).71 72
maḥ abbatan minnī (20, 39): Kūfa, Baṣra and Ḥ imṣ alone did not count it as a verse ending;
kay taqarra ʿaynuhā wa-lā taḥ zan (20, 40): Damascus (and in one narration, Ḥ imṣ) alone
counted it;
wa-fatannāka futūnan (20, 40): Baṣra, Damascus, and Ḥ imṣ alone counted it;
fī ahli Madyana (20, 40): Damascus (and in one narration, Ḥ imṣ) alone counted it;
wa-ṣtạ naʿtuka li-nafsī (20, 41): Kūfa, Damascus, and Ḥ imṣ alone counted it;
maʿanā banī Isrāʾīla (20, 47): Damascus and Ḥ imṣ alone counted it;
wa-laqad awḥ aynā ilā Mūsā (20, 77): Damascus (and in one narration, Ḥ imṣ) alone
counted it;
min al-yammi mā ġašiyahum (20, 78): Kūfa alone counted it;
waʿdan ḥ asanan (20, 86): Medina II (and in one narration, Damascus) alone counted it;
ġaḍbāna asifan (20, 86): Medina I and Mecca (and in one narration, Ḥ imṣ) alone counted it;
fa-kad̠ālika alqā l-sāmirī (20, 87): the second Medinan (and in some narrations, Damascus
and Ḥ imṣ) alone did not count it;
wa-ilāhu Mūsā (20, 88): Medina I and Mecca alone counted it;
fa-nasiya (20, 88): Medina I and Mecca alone did not count it;
ilayhim qawlan (20, 89): Medina II (and in some narrations, Damascus and Ḥ imṣ) alone
counted it;
id̠ raʾaytahum ḍallū (20, 92): Kūfa alone counted it;
qāʿan ṣafṣafan (20, 106): Medina I & II and Mecca alone did not count it;
minnī hudan (20, 123): Kūfa (and in some narrations, Ḥ imṣ) alone did not count it;
maʿīšatan ḍankan (20, 124): Ḥ imṣ alone counted it;
zahrata l-ḥ ayāti l-dunyā (20, 131): Kūfa (and in one narration, Ḥ imṣ) alone did not count it.
The sūra has 132 verses according to Baṣra, 134 according to Mecca and Medina I & II, 135
according to Kūfa, 138 according to Ḥ imṣ, and 140 according to Damascus. See al-Dānī,
al-Bayān, p. 88-108, 183-6; Spitaler, Verszählung, p. 44-7.
70
There are ten-verse separators at yā Mūsā (20, 11), al-ūlā (20, 21), azrī (20, 31), and yā
Mūsā (20, 40).
71
Due to the poor quality of the image, I cannot determine whether there is a verse separa-
tor at nusabbiḥ aka kat̠īran (20, 33). One would expect a verse ending there to match the one at
wa-nad̠kuraka kat̠īran (20, 34).
72
Where narrations about a city disagree, I disregard the city, yielding neither a match nor a
mismatch.

The presence of ten-verse markers at yā-Mūsā (20, 11), al-ūlā (20, 21), and
azrī (20, 31) to mark the tenth, twentieth, and thirtieth verses is compatible
with all the cities except Kūfa. However, the presence of a ten-verse separator
at yā-Mūsā (20, 40) to mark verse forty is compatible only with Kūfa, Mecca,
and Medina.
Sūra 31, 24-3473

There is a ten-verse marker at šakūr (31, 31), indicating the thirtieth verse.
This is compatible only with Mecca and Medina, ruling out Kūfa, Baṣra, and
Syria. There is no verse separator at lahu l-dīna (31, 32). This is compatible
with Kūfa, Medina, and Mecca, ruling out Baṣra and Syria. In sum, Baṣra and
Syria are ruled out twice, and Kūfa once. The verse divisions fit Mecca and
Medina.74
Sūra 32, 1-475

There is no separator at alif-lām-mīm. Kūfa alone puts a separator there, and
is thus excluded.
Sūra 59, 4-1076

There are no disputed verse endings in sūra 59, and the small leaf fragment
does not offer any variants.77
Synopsis
None of the disputed points distinguishing between Mecca, Medina I, and

Medina II happens to fall in the passages available for this study. But these
passages include plenty of disputed points that could decide between these
three Ḥ iǧāzī traditions taken together, on the one hand, and the cities outside
the Ḥ iǧāz, on the other hand. As one examines these instances, a striking pat-
tern emerges: There are fourteen cases in which the presence or absence of a verse
separator could have ruled out the Ḥ iǧāz in favor of some other locale(s), yet in
every single such case the Ḥ iǧāz passes the test. The probability of this pattern
occurring by chance is near zero. Ruled out conclusively are Kūfa, Baṣra,
Damascus, and Ḥ imṣ. These cities fail the test every single time the alterna-
tive is between them and the Ḥ iǧaz (six times for Kūfa, seven times for Baṣra,
73
Maṣāḥ if Ṣanʿāʾ, no. 4, p. 59 (English) = 44 (Arabic).
74
Al-Dānī, al-Bayān, p. 91, 106, and 206; Spitaler, Verszählung, p. 51.
75
Maṣāḥ if Ṣanʿāʾ, no. 4, p. 59 (English) = 44 (Arabic).
76
Image 152257-B in UNESCO CD.
77
See al-Dānī, al-Bayān, p. 243; Spitaler, Verszählung, p. 30.

and eight times for Syria). There is thus a stunning match between the qirāʾāt
literature and manuscript evidence.
III. The Lower Writing of Ṣanʿāʾ 1
Statement of the Problem
The lower writing is about as old as the parchment on which it appears (see
above, p. 354). And the words and phrases it conveys are at least as old as the
writing. The parchment, according to radiocarbon dating, has a 91.8 %
chance of dating from before ʿUt ̠mān’s death in AD 656, and a 95.5 % chance
of dating from before AD 661. It is almost certainly older than AD 671
(probability 98.8 %). Most likely, it was produced no more than 14 years
after the death of the Prophet Muḥammad: the probability of this is 75 %, or
a three-to-one likelihood. It is even slightly more likely than not that it was
produced within four years after the Prophet’s death (probability 56.2 %).
(For more details, see Table 1, above, p. 353.) What makes it even more likely
that Ṣanʿāʾ 1 was created in the first half of the seventh century is that after
ʿUt ̠mān’s standardization ca. AD 650, copying non-standard Qurʾāns became
less common. Those who defied the edict and held on to their Companion
codices were a minority. The standard version quickly became predominant,
as shown by the manuscript record and literary sources.
The lower writing, in any case, dates from the era of the Companions
of the Prophet. It was thus an early copy of a Companion C-1’s codex. So
far, this would make the C-1 textual tradition contemporaneous with the
ʿUt ̠mānic tradition, in the sense that they both date, at the latest, from the
era of the Companions. There is more to be said, however. In this essay I seek
to determine whether the wording of one codex can be determined to be
even older—in the sense of being either the source of the other or a signifi-
cantly more accurate copy of the common ancestor of both.
Two approaches to this question may be readily dismissed as callow. First,
it would be simpleminded to say, because C-1 may be labeled “pre-ʿUt ̠mānic”,
that its wording necessarily is older than that of the ʿUt ̠mānic text type. If by
“pre-ʿUt ̠mānic” one means “predating ʿUt ̠mān’s act of standardization”, then
the ʿUt ̠mānic text type may be “pre-ʿUt ̠mānic” too. The wording of the codi-
ces ʿUt ̠mān sent out may have predated the standardization, just as ʿUt ̠mān,
the people he set to the task, and the material they worked with existed
before AD 650. To ward off any prejudging of the matter, I call C-1 and other
Companion textual traditions “non-ʿUt ̠mānic” rather than “pre-ʿUt ̠mānic”.

Second, it is equally naive to say, because the upper writing came after the
lower one in this particular manuscript, that the wording of the ʿUt ̠mānic
Qurʾān must be later than that of C-1. That is tantamount to assuming that
when the upper writing appeared, it represented the totality of the ʿUt ̠mānic
tradition, meaning that the wording of ʿUt ̠mān’s codex did not exist before it
appeared on this particular manuscript. But the upper writing must have
been one of innumerable ʿUt ̠mānic manuscripts in circulation, a fact sup-
ported by the manuscript record, literary evidence, and common sense.78 It is
also unlikely that the lower writing represented the beginning and the end of
the C-1 textual tradition. In other words, one must distinguish between a
book as a physical object and a book as a text, for the wording is usually older
than the manuscript, and text types are almost always older than texts. That
one codex was produced after the other does not mean that its wording came
later. The upper layer of Ṣanʿāʾ 1, rather than capturing the genesis of the
ʿUt ̠mānic textual tradition, may represent its encroachment upon a parallel,
contemporaneous C-1 textual tradition. At least, that possibility cannot be
ruled out before analysis.
The fundamental question, therefore, remains wide open: what is the rela-
tionship between the text types of ʿUt ̠mān and C-1? Is it possible to determine
if one is older than the other? This question, of course, calls for textual criti-
cism and a model of textual transmission. I will argue that the model that
best explains the origin of the differences between the textual traditions of
C-1 and ʿUt ̠mān is dictation, though not word-by-word dictation. Imagine a
Companion of the Prophet doing his best to write down a Qurʾān (“Recita-
tion”) that was being recited faster than he could record with perfect accu-
racy. The Companion succeeded in getting most of the phrases right, thus
explaining the commonalities of the various Companion codices and ʿUt ̠mān’s
text type; yet, naturally, some differences arose. In due time, the Companion
C-1’s text was copied, and perhaps the copies were copied, creating a C-1
textual tradition and a family of closely related texts constituting the C-1 text
type, of which the lower writing of Ṣanʿāʾ 1 is one instance. In this scenario,
the repeated copyings did not necessarily involve dictation, but the C-1
archetype itself resulted from dictation.
The textual differences between C-1 and the standard Qurʾān are strikingly
similar to the variations that tended to arise in ḥ adīt̠ and āt̠ār texts in the late
second and early third century AH. By then, the use of writing was universal,
and standards of fidelity in transmission were approaching their peak. Yet
occasional differences arose between two versions of a text dictated by a com-
78
The extant manuscript fragments from the first century, which tend to be large, sturdy,
and expensive, must be but a fraction of early high-quality manuscripts; and expensive, high-
quality manuscripts must have been a small fraction of all manuscripts.

mon source. Such variations included not only those familiar from copyists’
errors—many such errors can occur during dictation, too—but also differ-
ences more typical of the oral and aural aspects of transmission by dictation.
Insights gathered from experience with this stage of ḥ adīt̠ transmission can be
brought to bear on the study of Companion Qurʾāns. Thus, for example, by
comparing the variants of a ḥ adīt̠ one finds that the most common words are
the most unstable, for they are the least memorable. In English, it would be
as if “say” and “tell” were more interchangeable than “perdure” and “endure”,
or “Mecca” and “Medina”. Moreover, in the ḥ adīt̠ literature, function mor-
phemes are the most variable. The English equivalents would be words like
“of ”, “or”, “so”, “by”, “to”, “with”, “it”, “and”, “then”, “some”. The most tell-
ing variations in ḥ adīt̠s, perhaps, involve the replacement of a word with a
similar-sounding word or changes of syntax that do not change the meaning.
Such changes occur less often in written transmission.79
The Impact of Semi-Orality: Patterns Governing the First Tier of Variations in

Transmission by Dictation
One sign of dictation—or dictation in which the outcome is not subse-

quently checked against the original—is that the changes are more frequent
than in written transmission.80 In both forms of transmission, the scribe’s
memory plays a role. It holds the words between the seeing/hearing and the
writing. However, memory usually plays a more profound role in dictation.
This is so for several reasons: Unless the dictation takes place slowly, the
scribe has to do two things at once: listen to new words and write down pre-
vious ones, making it easier to make mistakes. Moreover, in dictation, the
scribe’s glance does not fall on writing that can correct his memory. Further-
more, in dictation a scribe is more likely to copy sentence-by-sentence or
phrase-by-phrase than word-by-word, “which makes it easier to forget the exact
words”.81 The more rapidly the text is recited, the greater the strain is on the
scribe’s memory, and the larger and more numerous the resulting differences.
Aside from the larger number of variants, there is also a qualitative differ-
ence: each type of transmission generates certain types of changes in different
proportions. Some types of variants are more common in dictation and others
79
I base these conclusions on the study of numerous of āt ̠ār variants, in which I compared
isnāds with matns, as part of a current book project on early Islam.
80
For short, “dictation” in this essay generally refers to the type of dictation in which the
writing is not later corrected against the manuscript that was dictated. When the text is cor-
rected against the original, either visually or by reciting the text back, many of the changes are
eliminated.
81
The point about copying phrase-by-phrase was brought to my attention by Robert Waltz
(private communication, 2009).

more common in written transmission. For example, dictation eliminates

certain nonsense-generating errors that are due to the confusion of the eyes,
such as haplography, dittography, and errors due to the scribe’s eye leaping
from a word to a similar word elsewhere on the page; but nonsense-produc-
ing errors of the scribe’s hand can still happen.82
Changes arising from dictation tend to follow certain patterns that accord
with the limitations of human memory. It is convenient to refer to the most
basic of these changes as “first-tier”. First-tier variations are the most proba-
ble, and therefore most common, changes introduced by the limitations of
memory. Such changes can appear even through reasonably careful dictation.
However, changes exceed the first tier if the scribe exercises less care, if the
recitation takes place more rapidly, or if the time interval between the oral
statement and its writing is increased.83 In concrete terms, first-tier variations
may be discussed under five headings:
1. Changes of Minor Elements. The most commonly used elements of lan-

guage are what I refer to, strictly as a shorthand, as “minor”. These are the
easiest class of morphemes to add, drop, or substitute, for they are the least
memorable. Minor elements include function morphemes like particles, arti-
cles, conjunctions, prepositions, pronouns, and some prefixes and suffixes.
Minor elements also include a morpheme such as Allāh which occurs 2,689
times in the Qurʾān. Among the most variable elements are wa- and fa-. While
such changes do appear in written transmission, in dictation their absolute
numbers are greater.
Appendix 1 shows that in C-1 there are 35 minor differences with ʿUt ̠mān,
including 12 omissions, 19 substitutions (8 involving personal pronouns),
82
Some of the changes that are typical of dictation can be classified under what Metzger calls
“errors of the mind”. With regards to New Testament manuscripts, Metzger writes, “The category
of errors of the mind includes those variations which seem to have arisen while the copyist was
holding a clause or a sequence of letters in his (somewhat treacherous) memory between the glance
at the manuscript to be copied and the writing down of what he saw there” (Metzger, Text, p. 192).
He documents errors of the mind involving substitutions of synonyms, word order, transpositions,
and assimilation of parallels (ibid., p. 192-3). Note that while memory plays a role in both written
transmission and dictation, it plays a bigger role in the latter, which explains why errors of the mind
are more common in dictation.
83
The Ḥ adīt̠ literature furnishes ample evidence of the full spectrum of transmission behav-
iors, from purely oral to purely written. Changes in ḥ adīt̠s fit the profile of nearly complete
orality in part of the first century AH, reduce to high-tier changes by the end of that century
(signifying hybrid oral-written transmission), and then turn into patterns consistent with first-
tier or written transmission by the end of the second century AH. These shifts occur at differ-
ent rates in different circles of transmitters. It should also be noted that, in some stages, ḥ adīt̠
transmission could include hybrid oral-written forms of copying other than dictation, e.g. note
taking.

and four additions. It should be noted that there is a close match between
C-1 and ʿUt ̠mān, and that differences are the exception rather than the rule
even when it comes to minor elements. This leaves no doubt that transmis-
sion involved writing (as in dictation) rather than being purely oral.
The remaining principles, below, apply primarily to elements that are not
minor, thus labeled “major”. In dictation, major changes form a larger pro-
portion of variations than they do in written transmission. This is an obvious
consequence of the greater role played by memory in dictation.
2. Omissions of Major Elements. We more frequently forget a thing or
remember a thing differently than we “remember” something we have not
heard at all. Thus, words are dropped (or changed) more easily than they are
added. Furthermore, it is easier to forget a word if it is an item on a list. The
greater likelihood of omissions as compared to additions is a feature that dic-
tation shares with written transmission.84 In general, if a variant is minor, or
if it is an omission, then no special explanation is needed as to how it could
have come about. If it is both, its occurrence is even more understandable.
By contrast, if a variant is an addition, then it cannot be considered first-tier
a priori, and one must ask whether there are first-tier mechanisms that might
explain how it came about, mechanisms such as auto-contamination.
3. Auto-contamination. Auto-contamination refers to the influence, within
one textual tradition, of one part of the Qurʾān on another part. This is to
distinguish it from “cross-contamination”, which refers to one textual tradition,
84
In the case of dictation, at least, it is plausible that omissions occur more easily than addi-
tions. But omissions are somewhat more common even in other forms of transmission. New
Testament scholars used to believe that, ceteris paribus, the shorter of two readings is more likely
to be the original. This idea was challenged early in the twentieth century by the work of A.C.
Clark, who found that omissions are more common in the manuscripts of Latin and Greek clas-
sics. This led to divided opinions in the field. See Metzger, Text, p. 161-3; cf. Burnett Streeter,
The Four Gospels: A Study of Origins, 4th ed. rev., London, Macmillan, 1930, p. 131; following
Clark, Streeter states that “the error to which scribes were most prone was not interpolation but
accidental omission”.
More recently, studies of singular readings in papyri have confirmed repeatedly that omis-
sions were more common. See J.R. Royse, “Scribal Habits in the Transmission of New Testament
Texts”, in The Critical Study of Sacred Texts, ed. W.D. O’Flaherty, Berkeley, Graduate Theological
Union, 1979, p. 139-61; J.R. Royse, “Scribal Tendencies”, p. 239-52; Peter M. Head, “The
Habits of New Testament Copyists: Singular Readings in the Early Fragmentary Papyri of John”,
Biblica, 85 (2004), p. 399-408.
Singular readings are those that are unique to a manuscript. They subsume a disproportion-
ately large number of changes introduced by copyists and can thus be used to study the habits
and idiosyncrasies of scribes. For a discussion of methodology, see, in addition to the sources
cited above, E.C. Colwell, “Scribal Habits in Early Papyri: A Study in the Corruption of the
Text”, in The Bible in Modern Scholarship: Papers read at the 100th Meeting of the Society of Bibli-
cal Literature, ed. J.P. Hyatt, Nashville, Abingdon Press, 1965, p. 370-89.

say C-1, being affected at a point by another textual tradition, say ʿUt ̠mān or
Ibn Masʿūd. Auto-contamination takes two forms: (a) assimilation of paral-
lels and (b) assimilation of nearby terms.
(a) Assimilation of Parallels.85 This refers to a scribe changing the text,
making it more similar to a parallel passage in the work being transmitted
(auto-contamination) or in another branch of the textual tradition (cross-
contamination). It is the first of these, assimilation of parallels by way of
auto-contamination, that is relevant here.
The Qurʾān is—and calls itself—“self-similar” (Kor 39, 23). It is full of
repeated sentences and phrases that differ in one, two, or a few words. That is
why even today memorizers routinely find themselves adding or substituting
a word inadvertently if the added word appears in a similar sentence in a dif-
ferent verse. One’s knowledge of other passages shapes one’s memory of the
verse at hand, generating substitutions, additions, and deletions that hark to
the parallel. In dictation, hearing and memory are shaped by associations
formed by previous exposure to the Qurʾān. While assimilation of parallels
does occur in written transmission, it is more common in dictation because
of the greater role of memory.86
An omission or minor change does not require any additional explanation: it
can occur naturally as a first-tier transformation. If the variant is also a candidate
for auto-contamination, then its occurrence becomes even more understandable.
But additions of major elements are different. They do not occur as easily as
omissions. They cannot be considered as first-tier unless a first-tier mechanism,
such as the assimilation of parallels, can be invoked to explain them.
(b) Assimilation of Nearby Terms. It is more likely for a word to be used by
mistake at a certain point if it is used in a nearby passage. A word is on the
scribe’s mind if he heard it a moment ago or if he expects to hear it soon due
to prior familiarity with the passage at hand. Such a word can insinuate itself
into the writing.87
85
Assimilation of parallels, also called “harmonization”, is a very common source of change
in New Testament manuscripts. See Burnett Streeter, The Four Gospels, p. 139-44; E.C. Colwell,
“Scribal Habits”, p. 377-8; Metzger, Text, p. 193, 197-9; Peter M. Head, “Observations on Early
Papyri of the Synoptic Gospels, Especially on the ‘Scribal Habits,’” Biblica, 71 (1990), p. 240-7;
Head, “Habits”.
86
If the scribe taking dictation subsequently compares the result with the written source and
then makes corrections, the results will look different: there will be far fewer changes than in
dictation alone. The extent to which this relatively careful variety of dictation may have played
a role in the transmission of the ʿUt ̠mānic Qurʾān is worth investigating.
87
Assimilation of nearby terms, or “harmonization to the immediate context”, is an important
category of scribal error in New Testament manuscripts. See Colwell, “Scribal Habits”, p. 377-8;
Royse, “Scribal Tendencies”, p. 246; Head, “Observations”; Head, “Habits”.

4. Phonetic Conservation in Major Substitutions. Because of the aural

dimension, if a major element changes, more often than not it will sound like
the original. For example, the same words may appear in a different order, or
a verb may be changed to another verb that has the same root. Some types of
phonetic conservation are listed in the left column of Table 6:
Table 6. Types of Phonetic Conservation of Major Elements, with Examples.

Type of Difference Verse ʿUt ̠mān C-1
word order 5, 41 wa-lahum fī wa-fī l-āḫ irati
(transposition) l-āḫ irati lahum
another verb, same root 2, 196 istaysara tayassara
verb conjugated for 15, 65 tuʾmarūna yaʾmurūna
different person
different form of plural 5, 43 al-nabiyyūna al-anbiyāʾu
singular vs. plural 2, 219 al-āyāti al-āyati
verb tense or mood 5, 52 fa-yuṣbiḥ ū fa-yuṣbiḥ ūna
active vs. passive 5, 47 anzala llāhu fīhi unzila ʿalayhim fīhi
Of the 25 cases of major substitution, i.e. cases where C-1 and ʿUt ̠mān use
different words or phrases, 18 involve similar-sounding variants, while only 7
involve variants that do not sound similar. Appendix 1 lists the differences (in
major elements) between C-1 and ʿUt ̠mān that are characterized by phonetic
conservation, while Appendix 2 (c) lists those that are not.
5. Common or Frequent Terms. First-tier transmission generates no or very
few additions that cannot be accounted for by auto-contamination. If such an
addition does occur, however, it will have to be a frequently-used word. In
any case, first-tier changes would not include the addition of a word that is
unexpected and uncommon.
First-tier transmission generates, likewise, few substitutions that cannot be
accounted for by either phonetic conservation or auto-contamination. Yet
such substitutions can occur more easily than additions, since it is easier to
misremember a word than “remember” a word that has not been said. If this
happens, memory usually substitutes a familiar word with another word that
is familiar or frequent, or a frequent word with another frequent word. A
word is “familiar” if it is used elsewhere in the Qurʾān. (Less commonly does
memory replace a rare word with a frequent word, and even less commonly
does it replace a familiar word with a rare word. These last two transforma-
tions are not first-tier.)

With less exactness in transmission, e.g. due to faster dictation or purely

oral transmission, some of the first-tier tendencies are progressively relaxed. It
is convenient to refer to the less precise forms of transmission as “higher-tier”.
Higher-tier transmission practices still possess the hallmarks of dictation
listed above, but they also show additional types of changes. For example,
phonetic conservation is obeyed less frequently.
The most important observation one can make concerning C-1 is that its
differences with the standard Qurʾān generally do not exceed first-tier varia-
tions. This confirms that the appropriate model is one of transmission by dic-
tation rather than pure orality. Dictation explains not only why C-1 and
ʿUt ̠mān are generally the same, even when it comes to minor elements, but
also how the differences may have arisen. In particular, the role of aurality is
evident in the large proportion of variations in minor elements, phonetic
conservation of major elements, and the significant role of auto-contamination.
To say that the differences between ʿUt ̠mān and C-1 are first-tier begs the
question of directionality. Some of the five tendencies listed above are asym-
metric, as they can accommodate change in one direction more easily than
the other. These can be used to see if it is easier to get, by dictation, from
codex A to codex B or the other way around. I will show that going from
ʿUt ̠mān to C-1 would be first-tier, but not the other way around. This is the
second of two methods applied here to answer the question of priority.
The first method applied here, stemmatics, constructs a stemma, i.e. a fam-
ily tree, using patterns of agreement and disagreement among text types. To
do so, one needs to compare at least three texts. Therefore, I bring into play
the codex of Ibn Masʿūd, a non-ʿUt ̠mānic text which, while not attested in
manuscripts, is described in literary sources. Comparisons will be made with
other Companion codices where data is available. This exercise presupposes
that there is a significant kernel of truth in what the literary sources say about
the codex of Ibn Masʿūd. This is a point that requires discussion before pro-
ceeding further.
The Codex of Ibn Masʿūd and other non-ʿUt̠mānic Qurʾāns
Literary sources speak of codices of Companions of the Prophet such as Ibn

Masʿūd (popular in Kūfa),88 Ubayy b. Kaʿb (popular in Syria and Baṣra),89
Abū Mūsā l-Ašʿarī (Baṣra and Yemen),90 and others. The codex of ʿAbd Allāh
88
For a discussion of the muṣḥ af of Ibn Masʿūd, see Rāmyār, Tārīḫ-i Qurʾān, p. 353-64.
89
For a discussion of the muṣḥ af of Ubayy b. Kaʿb, see Rāmyār, Tārīḫ-i Qurʾān, p. 340-53.
90
For a discussion of the muṣḥ af of Abū Mūsā l-Ašʿarī, see Rāmyār, Tārīḫ-i Qurʾān,
p. 380-4.

b. Masʿūd, uniquely among the Companion codices, reportedly did not have
the last two sūras of the standard Qurʾān, which together amount to forty-
three words, which happens to be exactly the number of words found in this
sentence.91 In the codices of Ibn Masʿūd and Ubayy b. Kaʿb, the sūras are
arranged in different orders relative to one another; but, within the sūras, the
codices generally have the same arrangement of sentences as the standard
Qurʾān. There were, however, occasional differences at the level of words and
phrases. These differences have been reported in a variety of Muslim tradi-
tions from the first two centuries AH that are preserved in later Muslim liter-
ary sources. It would be naive to accept all such reports as reliable descriptions
of Companion codices; but C-1 proves it equally unwise to dismiss them all as
unreliable.
Among the reports about Companion codices, one tradition stands out
because of its methodical nature. It is a report of the prominent second-cen-
tury Kūfan traditionist and Qurʾān-reciter, Sulaymān b. Mihrān al-Aʿmaš
(d. 147/765, Kūfa). He lists Ibn Masʿūd’s “reading” at over 150 points.92 The
list includes many single-letter variants, such as the use of wa- (“and”) versus
fa- (punctuation). It also includes more important variations, such as substi-
tutions, transpositions, additions, and omissions of words, although most of
these make little difference to the meaning. The routine inclusion of minute
variants, combined with the systematic nature of the list, creates the impres-
sion that the bulk of it is based on an actual codex. Furthermore, on the face
of it very few variants appear to reflect a later development or to have been
generated to further a theological or sectarian agenda.93
91
It is also said that the codex of Ibn Masʿūd lacked the first sūra (which is twenty-five words
long). This may have been the case. But al-Aʿẓamī says that this is contradicted by the fact that
Ibn Masʿūd’s reading includes variants for this sūra. He also points out that Ibn al-Nadīm saw a
muṣḥ af ascribed to Ibn Masʿūd that had the sūra (al-Aʿẓamī, The History of the Qurʾānic Text, p.
235-6; cf. Ibn al-Nadīm, al-Fihrist, p. 29, where he says that the manuscript was from AH
200). Note that al-Aʿmaš, at least, does not report variants of Ibn Masʿūd for the first sūra. For
references to the primary sources with these reports about the first and the last sūras, see Modar-
ressi, “Early Debates”, p. 13, footnote 48; al-Aʿẓamī, The History of the Qurʾānic Text, p. 236.
Al-Aʿẓamī dismisses the reports he quotes as self-contradictory (ibid.); but a plausible harmoni-
zation would not be difficult to imagine. In the final analysis, however, it must be admitted that
the reports are uncertain.
92
Ibn Abī Dāwūd, al-Maṣāḥ if, p. 57-73.
93
One variant of Ibn Masʿūd as reported by al-Aʿmaš gives me pause: in Kor 3, 7, in lieu of
the standard phrase, it reads wa-in ḥ aqīqatu taʾwīlihi illā ʿinda llāhi wa-l-rāsiḫūna fī l-ʿilmi
yaqūlūna āmannā bihi (Ibn Abī Dāwūd, al-Maṣāḥ if, p. 59). It is possible that this is a deliberate
change intended to remove the ambiguity of the standard version. A critic might object that this is
a conspiracy theory. But it is not conspiracy-minded to allow for the possibility of deliberate change
in an individual report (which may or may not have been accepted by others) in the way it

Among the variants ascribed to Ibn Masʿūd and Ubayy b. Kaʿb, several are
anomalous: they add large new clauses or phrases absent in the ʿUt ̠mānic text
that are not explainable in terms of the aural dimension of dictation—and
these happen to explicitly endorse the ʿAlīd cause.94 It would be natural to
dismiss these as tendentious interpolations; but to generalize this negative
judgment to the totality of reported variants would be to throw the baby out
with the bathwater. Rather, one expects some of the sources of non-ʿUt ̠mānic
readings to be more trustworthy than others. In particular, as explained in
the last paragraph, the Ibn Masʿūd variants given by al-Aʿmaš do not seem to
be tendentious. Nevertheless, it remains true that al-Aʿmaš does not disclose
his sources,95 and research clearly is needed into the sources and transmission
of his and other non-ʿUt ̠mānic variants.
A key point about al-Aʿmaš’s version of Ibn Masʿūd’s codex is that more
often than not its variants are related to the ʿUt ̠mānic text through first-tier
transformations. Minor elements form a large category, and, as shown in
Table 7, the use of synonyms and the tendency to conserve major elements
phonetically are in evidence. There is also evidence of auto-contamination,
most noticeably in the form of assimilation of nearby terms.
Comparing Table 7 (on Ibn Masʿūd) with Table 6 (on C-1) or Appendix 1
reveals striking similarities. In many cases, Ibn Masʿūd (known via al-Aʿmaš)
relates to ʿUt ̠mān in ways that are familiar from the ways in which C-1 relates
to ʿUt ̠mān. That lends credence to the historicity of the Ibn Masʿūd codex
and suggests that it indeed possessed many of the variants ascribed to it. By
this, I do not mean that C-1 and Ibn Masʿūd share actual variants, which
they do in a few cases, but that they share some of the same types of variants.
In general, every type of variant found in C-1 is found also in Ibn Masʿūd.
However, Ibn Masʿūd also has some higher-tier types not found in C-1. (For
these types, one cannot yet claim corroboration from manuscripts, warrant-
ing some caution regarding their authenticity.) This means that, on the
whole, the standard Qurʾān is closer to C-1 than it is to the codex of Ibn
Masʿūd—qualitatively that is, but not in terms of the quantity of variations.
It is not just by way of verbal variants that C-1 supports the historicity of
non-ʿUt ̠mānic codices and the reliability of some of the reports about them;
there are also the similarities in the order in which some sūras appear in C-1
and the sūra sequences reported for the codices of Ibn Masʿūd and Ubayy b.
Kaʿb, as shown in Table 8.
would be to speak of deliberate changes accepted by entire communities in an already well-known

divine text.
94
For the details, see Modarressi, “Early Debates”, p. 25-6.
95
Al-Aʿẓamī, The History of the Qurʾānic Text, p. 234.

Table 7. Examples of the phonetic conservation of major elements between

Ibn Masʿūd and ʿUt̠mān. Compare to Table 6.
Type of difference Verse ʿUt ̠mān Ibn Masʿūd (via al-Aʿmaš)

word order 3, 156 wa-llāhu bi-mā wa-llāhu baṣīrun bi-mā
(transposition) taʿmalūna baṣīrun taʿmalūna
synonymous verb, 2, 202 kasabū iktasabū
same root
verb conjugated for 2, 83 taʿbudūna yaʿbudūna
different person
different form of 13, 42 l-kuffāru l-kāfirūna
plural
singular vs. plural 34, 37 l-ġurufāti l-ġurfati
tense or mood 17, 44 tusabbiḥ u sabbaḥ at
active vs. passive 32, 17 qāla aʿlamu qīla aʿlamu
Table 8. The sūra order in C-1 compared to those reported for Ibn Masʿūd
and Ubayy b. Kaʿb. The numbers given represent the ʿUt ̠mānic order.
C-1 Ibn Masʿūd Ubayy b. Kaʿb
63, 62, 89, 90 The Fihrist list:96 The Fihrist list:98
63, 62, . . . [twenty-nine 63, 62, 65, 89, . . .
intervening sūras], 89, 85, sūra 90 does not appear on
84, 96, 90 the list.
The Itqān list, quoting The Itqān list:99
al-Aʿmaš:97
63, 62, . . . [twenty-seven 63, 62, 66, 89, 90
intervening sūras], 89, 85,
84, 96, 90
96
Ibn al-Nadīm, al-Fihrist, p. 29; Rāmyār, Tārīḫ-i Qurʾān, p. 356-9; Neal Robinson, Disco-
vering the Qur’an: A Contemporary Approach to a Veiled Text, 2nd ed., London, SCM Press,
1996, p. 263-6; Arthur Jeffery, Materials for the History of the Text of the Qurʾān: The Old Codi-
ces, Leiden, Brill, 1937, p. 22.
97
Al-Suyūṭī, al-Itqān fī ʿulūm al-Qurʾān, n.p., Dār al-fikr, 1416/1996, I, p. 176; Rāmyār,
Tārīḫ-i Qurʾān, p. 359-61; Robinson, Discovering the Qurʾān, p. 263-6; Jeffery, Materials, p. 23.
98
Ibn al-Nadīm, al-Fihrist, p. 29-30; Rāmyār, Tārīḫ-i Qurʾān, p. 344-5; Jeffery, Materials, p. 115.
99
Al-Suyūṭī, al-Itqān, I, p. 175-6; Rāmyār, Tārīḫ-i Qurʾān, p. 346; Jeffery, Materials, p. 115.

One may make two observations about C-1’s sūra ordering. First, by virtue of
the 63-62 sequence, it can be considered closer to the Ibn Masʿūd and Ubayy
b. Kaʿb codices than to ʿUt ̠mān. Second, it is over-all much nearer to Ubayy
b. Kaʿb than it is to Ibn Masʿūd.
Which Text Type is Older? Method I: Stemmatics
I shall now attempt to construct the family tree, or stemma, of three text
types, those of Ibn Masʿūd, ʿUt ̠mān, and C-1, based on the patterns of agree-
ment and disagreement among them. It is impressive that Ibn Masʿūd’s codex
agrees with C-1 several times, as shown in Table 4; but far more striking is
how exceptional such agreements are. Usually, where there is a difference
between ʿUt ̠mān and Ibn Masʿūd (or, for that matter, between ʿUt ̠mān and
any other Companion codex), C-1 sides with ʿUt ̠mān. In other words, in
almost every case in which there is a disagreement between the three, ʿUt ̠mān
is in the majority: either it is ʿUt ̠mān and C-1 against Ibn Masʿūd, or it is
ʿUt ̠mān and Ibn Masʿūd against C-1. ʿUt ̠mān is rarely alone: that is, seldom
do Ibn Masʿūd and C-1 agree against ʿUt ̠mān. Therefore, ʿUt ̠mān occupies a
central position, which may be represented as follows:
C-1 — ʿUt ̠mān — Ibn Masʿūd
Depending on which of its three nodes one holds up or pulls down, this pat-
tern yields different stemmata. The constraint imposed by this pattern cuts
down the number of feasible stemmata of the text types including a proto-
type from sixteen to the six shown in the following diagram, labeled (a)
through (f ). The case in which ʿUt ̠mān is pulled down and the others are held
up is shown later in (g).100
Here, P stands for the “Prototype”—as I shall argue later, to be identified
with the Prophet. IM, ʿUt ̠, and C-1 refer respectively to the text types (or
textual traditions) of Ibn Masʿūd, ʿUt ̠mān, and C-1. If one does not posit a
prototype, one obtains only stemmata (a), (b), and (e), with P removed. This
set of stemmata is what the verbal features yield if the only information that
is taken into account is the pattern of verbal agreement and disagreement
between the codices and if the exceptional cases are set aside. The question is
whether there are other grounds for preferring one stemma over another. The
following paragraphs show that there are reasons to favor stemmata (e) and
100
On the construction of stemmas, see L.D. Reynolds and N.G. Wilson, Scribes & Scholars:
A Guide to the Transmission of Greek & Latin Literature, 3rd ed., Oxford, Clarendon Press, 1991,
p. 211-3. I owe this reference to Michael Cook.

P P P P
C-1 IM IM ʿUt ̱ C-1 ʿUt ̱
ʿUt ̱ ʿUt ̱
IM C-1 C-1 IM
(a) (b) (c) (d)
P P
ʿUt ̱
IM C-1 IM ʿUt ̱ C-1

(e) (f)
(f ), the latter being more probable. Note that under (f ), the fact that ʿUt ̠ is in
the majority entails that it is more reliable than IM and C-1.
All the stemmata would have to describe acts of transmission before the
promulgation of the ʿUt ̠mānic Qurʾān. For by the time ʿUt ̠mān had Qurʾāns
copied and sent to the cities, copies conformed very closely to the wording of
their sources, differing from them typically only in units of one or two
letters;101 and, thereafter, as manuscript evidence shows, changes were of the
101
Cook, “The Stemma of the Regional Codices of the Koran”, p. 90-1. For a discussion, see
the section above on the upper text.

limited type expected of written transmission. The post-ʿUt ̠mān range of ver-
bal variations was thus relatively small. While this observation does not point
to a stemma, it highlights a crucial fact that helps narrow down the choices:
there was a drastic transition from a mode of transmission generating first-
tier changes due to semi-orality to a mode of transmission with relatively
small changes that are characteristic of copying from manuscript. This radical
shift took place in the twenty years or so before ʿUt ̠mān’s standardization, not
a long interval. Furthermore, it took place either at a specific point in time
(e.g. as soon as P had dictated the text in scenario (f )) or gradually. Now, if
one were to accept (a), (b), (c), or (d), that would make the variants the prod-
uct of two or three successive generations of textual reproduction, and one
would expect the abovementioned shift (in reproductive fidelity) to be
reflected in the codices by virtue of the intermediate codex, ʿUt ̠, being ver-
bally much closer to the youngest codex than the oldest.102 The data does not
quite fit such a pattern. ʿUt ̠ is slightly closer to C-1 in terms of the quality of
variants but closer to IM in terms of their quantity. It also counts against
these stemmata that IM and C-1 occasionally agree against ʿUt ̠. These consid-
erations leave (f ) as the plausible choice, with (e) being the second best. The
stemma (g), to be discussed later below, remains as plausible as (f ).
The fact that fidelity in transmission increased rapidly within a couple of
decades shows that the biggest changes must have been introduced in the first
stage of transmission. (This reminds one of other situations where the largest
variations are the oldest, such as in the Ḥ adīt̠ literature or the transmission of
New Testament manuscripts.)103
The choice (f ) would also follow from a scenario in which the text was
originally dictated, and thereafter it was copied from manuscripts or was dic-
tated and then corrected against the written source. On this scenario, the dif-
ferences that are due to orality must go back to the original moment of
102
If either (a) or (b) were the true stemma, then the wording of either C-1 or IM would
be much more similar to ʿUt ̠ than to the other. If IM were the one much more dissimilar to
ʿUt ̠, we would prefer (b) to (a); but if C-1 were the one much more different, we would prefer
(a) to (b).
103
Spencer et al. lend credence to the “ideas that most variants arose early in the history of
the Greek New Testament, . . . and that later copies introduced fewer variants” (Matthew Spen-
cer, Klaus Wachtel, and Christopher Howe, “Representing Multiple Pathways of Textual Flow
in the Greek Manuscripts of the Letter of James Using Reduced Median Networks”, Computers
and the Humanities, 38 (2004), p. 1-14). This comparison, however, should not blind one to
the profound differences between the New Testament and the Qurʾān, including the fact that
the Qurʾān was already scripture at its inception and constantly identifies itself as such, and the
fact that Islam became a state religion already as the Qurʾān was being revealed. These factors
explain why the text of the Qurʾān was stabilized so much more rapidly than that of the New
Testament.

dictation, and thus only (f ) fits the data.104 The Qurʾān perhaps lends some
support to this scenario by its self-description as a written text akin to other
divine scriptures. It identifies itself as having been divinely revealed to the
Prophet and as having taken the form of an oral utterance by him (Kor 75,
16-8; 98, 2-3) that was put in writing (Kor 25, 5; 98, 2-3),105 revelations that
Muslims were expected to recite as frequently as possible (Kor 73, 20), a
scripture akin to other holy books, the Torah and the Gospels (Kor 9, 111).
One may read these Qurʾānic statements as being consistent with the idea that
the Prophet dictated the Qurʾān to a number of scribes and that it was there-
after transmitted in writing, like the Torah and the Gospels. Note, however,
that even without this scenario stemma (f ) remains plausible.
In short, the options are (e) and (f )—stemma (f ) being more likely.106 The
former (e) would make ʿUt ̠mān the common source of Ibn Masʿūd and C-1.
The latter (f ) would make the three text types independent reproductions of
104
The scenario in which the Qurʾān was dictated by the Prophet and thereafter copied by
written transmission would be plausible if there were no interdiction on written transmission.
And indeed there is no sign of any such interdiction for the scripture. On the other hand, atti-
tudes toward the writing of non-scriptural religious literature were ambivalent. For the state of
this controversy in the second century AH, see Michael Cook, “The Opponents of the Writing
of Tradition in Early Islam”, Arabica, 44 (1997), p. 437-530.
105
Rāmyār cites a number of verses to support the idea that the Qurʾān was put in writing
by Muḥammad. The most definitive one is Kor 25, 5. There is also Kor 98, 2-3. The Qurʾān’s
mention of teaching by the pen (Kor 96, 1-5) may be an allusion to divine scriptures and the
role of writing in their transmission. The verses at Kor 74, 52 and 6, 91 are highly suggestive.
The cases of Kor 80, 11-6; 85, 21-2, and 56, 77-80 are relevant, although, as Rāmyār acknowl-
edges, less conclusive. Rāmyār also gives a variety of other Qurʾānic references to writing, copy-
ing, and their accoutrements such as paper and parchment. He also highlights the Qurʾān’s
references to the Jewish Bible as a physical object. See Rāmyār, Tārīḫ-i Qurʾān, p. 257-8, 275-9.
It would be of interest to know how Jewish scriptures were transmitted in Arabia on the eve of
Islam. Cook has shown that some early Muslim ideas about written and oral scriptures reso-
nated with Jewish notions (Cook, “Opponents”, cited in the last footnote).
106
One may also attempt to choose among the stemmata using the data of the sūra ordering
of the codices. The fact that the orderings of Ibn Masʿūd and C-1 are closer to each other than
they are to ʿUt ̠mān would exclude all stemmas except (e) and (f ). If we went with (f ), then one
of the following three scenarios would explain the similarity of IM and C-1: (i) they reflected
P’s sūra ordering better than ʿUt ̠, (ii) one influenced the other (“contamination”), or (iii) they
independently used a similar principle to order the sūras. If we went with (e), then i would
become moot and we would have to choose between ii and iii.
Note that if ii or iii are true, however, then all the stemmata in the diagram would regain
plausibility and sūra ordering would not help us choose between them. I place this discussion
in a footnote because it is very easy to imagine ii and iii as valid scenarios for sūra ordering,
especially if, as it seems to be the case, at the time of the Prophet many sūras were free-standing,
not yet rigidly ordered in a fixed sequence. Scenario iii is especially plausible since the sūra
sequences ascribed to Ibn Masʿūd and Ubayy b. Kaʿb appear intended to arrange most sūras in
order of decreasing length (Robinson, Discovering the Qurʾān, p. 263-6).

(IM) (C-1) C-2 … C-n
ʿUt ̱
(g)
the prototype; however, ʿUt ̠mān’s version would be the most reliable of the
three, for in cases of disagreement it is generally in the majority. Only excep-
tionally do Ibn Masʿūd and C-1 agree against ʿUt ̠mān. These commonalities
are due to “cross-contamination” (one textual tradition influencing the
other), or are independent “convergent adaptations”, or indeed go back to
the prototype. As discussed further below, in this last case the possibility
remains that, as some Muslim traditions maintain, the Prophet had dissemi-
nated revised versions of the prototype.
The analysis so far has treated the ʿUt ̠mānic codex as a package in which
different parts have the same textual history. If, however, one entertains the
possibility that ʿUt ̠mān is a hybrid text based on a number of Companion
codices, C-2, . . ., C-n, and possibly IM and C-1, then the stemma will be (g).
In this scenario, the fact that ʿUt ̠mān’s version gives the majority reading
means that it emerged as part of an attempt to find consensus. Where two
codices disagreed, ʿUt ̠mān’s committee would have sought independent cor-
roboration by other Companions. This happens to be more or less how a
modern scholar would set out to reconstruct the prototype if he had access to
a number of independent copies of a common source. And, indeed, such a
procedure can be expected to have produced a better text than any of the
individual codices (IM, C-1, C-2, etc.), i.e. one that better preserved P’s
wording. Stemma (g) and the ideas in this paragraph are due to Michael
Cook.107
107
Michael Cook, “Concluding Remarks”, Colloquium on the Evidence for the Early His-
tory of the Qurʾān, Stanford University, July 31, 2009.

This scenario does not ascribe to the committee an unreasonably high

degree of sophistication. The practice of verifying a report by seeking corrob-
oration is a matter of common sense, and it is mentioned in the Qurʾān (Kor
24, 12-3; 2, 282). And while for the purpose of this essay I suspend judg-
ment on the accuracy of reports of individuals, it is worthwhile noting that
two reports, with independent isnāds, traceable to two different scribes in
ʿUt ̠mān’s team state that in cases of disagreement over wording, the disputed
point was postponed or the spot was left blank so that it could be dealt with
later. If true, that would support hybridity.108
To sum up, the evidence of the palimpsest inspires confidence that literary
sources carry partly valid information about non-ʿUt ̠mānic codices. In par-
ticular, I provisionally accept al-Aʿmaš’s description of the codex of Ibn Masʿūd
(above, p. 391).109 With three codices available, it is natural to seek to deter-
mine genetic relationships. ʿUt ̠mān occupies a central position between the
other two. This indicates either that ʿUt ̠mān is the more faithful copy of a
common source, the prototype, or that it is a hybrid reflecting the majority
reading of Companion codices. Chronological constraints help identify the
prototype with the Prophet’s own recension, and the Prophet as the person
who dictated it—an argument that will be spelled out in a later section
below.
Which Text Type is Older? Method II: Internal Evidence and Polarity Analysis
The application of stemmatics above relies on certain propositions: that we

know Ibn Masʿūd’s codex and that the biggest verbal changes were introduced
at an early stage in transmission. These may be excellent premises; but it
would be gratifying if there were an additional argument that did not rely
on them.
I turn to analyzing the internal evidence. Do ʿUt ̠mān and C-1 have the
same character relative to each other, or is there an asymmetry that shows
that one of them has the older wording? To answer this question, one should
ask which is easier to imagine: scribal mistakes that transform ʿUt ̠mān’s word-
ing to that of C-1, or scribal mistakes that change C-1 to the ʿUt ̠mānic
108
Ibn Abī Dāwūd, al-Maṣāḥ if, p. 21-2 and 25-6; cf. Rāmyār, Tārīḫ-i Qurʾān, p. 426, where
the author cites additional sources for one of these reports.
109
Although I used the codex of Ibn Masʿūd as reported by al-Aʿmaš in attempting to con-
struct the stemma, and although I have attempted to show that al-Aʿmaš’s report is largely cred-
ible, it is important to note that the validity of my argument is not sensitive to the details of the
Aʿmaš report. If instead of that report, I used all or some of the other reports about the codex of
Ibn Masʿūd, Ubayy b. Kaʿb, or any other Companion, then ʿUt ̠mān would still be in the major-
ity, and the stemmata would be the same.

version? In other words, is the transformation of one codex to the other more
consistent with first-tier changes (p. 386-9) than the other way around? I will
show that, indeed, the transformation of ʿUt ̠mān to C-1 can be explained in
terms of first-tier changes, but the transformation of C-1 to ʿUt ̠mān cannot.110
This is a highly non-trivial observation. Suppose that the dictation of a
common source generated two codices, A and B. Even if A and B were both
first-tier copies of their common source, in general the relationship between
A and B would not be reducible to the first tier, i.e. one could not show that
one can get from A to B (or from B to A) with first-tier transformations. (For
example, A could omit some words and B some other words, so each would
have major terms lacking in the other which could not be explained by auto-
contamination, etc., making it impossible to show that you can get from one
codex to the other by first-tier transformations.) So, the fact that one can get
from ʿUt ̠mān to C-1 by first-tier transformations means either that ʿUt ̠mān
(i.e. its wording) is the source of C-1 or that it is an accurate version of their
common source.
Let us now see if it is as easy to get from one codex to the other as the
other way around. The most useful set of differences for the analysis are cases
in which one codex has a major element in a verse that the other codex does
not. I call these major pluses. Consider, first, those cases in which a passage
in C-1 has a word not found in the same passage in ʿUt ̠mān. Let us label such
extra terms as pluses of C-1. Similarly, there are cases in which a passage in
ʿUt ̠mān has an extra term not found in C-1’s version of that passage. These
are the pluses of ʿUt ̠mān.
(1) Omissions. It is easier for a scribe taking dictation to drop and change
major words he has heard than to add major words he has not heard. ʿUt ̠mān
has fourteen major pluses while C-1 has ten (see Appendix 2). As for minor
pluses, ʿUt ̠mān has twelve while C-1 has only four (see Appendix 1). The
total is twenty-six pluses in ʿUt ̠mān versus fourteen in C-1. This suggests that
through dictation it is easier to get from ʿUt ̠mān (or from a prototype that
ʿUt ̠mān represents faithfully) to C-1 than the other way around. However,
numbers alone are not decisive. It is the nature of the pluses that clinches the
argument: it will be noted below (3) that all the major pluses of C-1 can be
ascribed to other first-tier mechanisms such as auto-contamination, but that
cannot be done for all the major pluses of ʿUt ̠mān.
(2) Omissions in Lists. Keeping several items in mind simultaneously can
be a challenge. If a scribe who takes dictation forgets a word, it is easy to see
110
In this entire section, by “transforming codex A to codex B”, I mean “getting from the
wording of codex A to the wording of codex B”.

how this could happen if the word is an item on a list. It is, therefore, signifi-
cant that of the pluses of ʿUt ̠mān, two appear in lists of three items apiece. In
C-1, these lists appear with only two items. For the details, see Appendix 2,
section (a) (“Omissions”), Kor 2, 196 and Kor 2, 218. It is easier to see how
one could get from ʿUt ̠mān (or its prototype) to C-1 than the other way
around.
(3) Major Pluses. The most important pattern is the way in which the
pluses in C-1 differ from those in ʿUt ̠mān. I begin with C-1. There are ten
major pluses in C-1, and these are listed and annotated in Appendix 2, sec-
tion (b) (“Additions”). Prior to analysis, these words can be considered either
additions of C-1 or omissions of ʿUt ̠mān.
Their most telling attribute is that every one of them, without exception,
appears in a parallel passage elsewhere in the standard Qurʾān that has that
word. (A passage is a “parallel” if it, too, contains some of the words or
phrases surrounding the plus.) Take Kor 2, 193. Here, C-1 has “And fight
them until persecution is no more and religion is all for God.” The word “all”
is a plus of C-1; ʿUt ̠mān does not have it. Yet ʿUt ̠mān has this exact sentence,
including “all”, in another verse, namely Kor 8, 39. The extra “all” of C-1 at
Kor 2, 193, therefore, is a candidate for assimilation of the parallel found in
Kor 8, 39. That is to say, the word “all” might have occurred to C-1 because
of the association with the other verse.
It is easy to see how the nine major pluses of C-1 could be secondary to
the ʿUt ̠mānic version: they are candidates for auto-contamination. But can
one make the same argument, in reverse, for the pluses of ʿUt ̠mān, to show
that ʿUt ̠mān could be secondary to C-1? Is the situation symmetric? No, it
is not.
It is different with the pluses of ʿUt ̠mān. There are fourteen major pluses in
ʿUt ̠mān—see Appendix 2, section (a) (“Omissions”). Among them are five
that occur in sentences that are not attested or paralleled anywhere else in the
Qurʾān. Nor do these five terms show up anywhere else with phrases that sur-
round them in the verse at hand. Therefore, they could not have been gener-
ated by auto-contamination. These “orphans” are as follows: Kor 2, 196 (aw
ṣadaqatin), 2, 217 (wa-kufrun bihi), 2, 222 (fa-ʿtazilū), Kor 5, 42 ( fa-in
ǧāʾūka), and 63, 1 (qālū); although one perhaps should discard the last two
cases—Kor 5, 42 because of its proximity to id̠ā ǧāʾaka in Kor 5, 48 (making
it a candidate for assimilation of nearby terms), and Kor 63, 1 because qālū is
a very common term in the Qurʾān.
It is worth giving a systematic and precise paraphrase of the argument.
First consider the pluses of C-1. If the wording of C-1 is older than that of
ʿUtmān, this would mean that ʿUtmān omitted these terms. Omissions are

consistent with first-tier transmission. So, it would be easy to get from C-1 to
ʿUtmān through dictation. On the other hand, if the wording of ʿUtmān is
older that that of C-1, the pluses of C-1 would be items added by C-1. Addi-
tions cannot be first-tier changes unless they are candidates for auto-contami-
nation. Indeed, all the pluses of C-1 can be explained as possible cases of
auto-contamination: these terms, without exception, are found in parallel
passages. Thus, it is equally easy to imagine ʿUtmān as the older text. In sum,
the pluses of C-1 do not favor one codex over the other. This is because
depending on which text is older, the pluses of C-1 are either omissions of
ʿUtmān or auto-contaminations of C-1.
The situation is different with the pluses of ʿUt ̠mān. If ʿUt ̠mān has the
older wording, these would be omissions of C-1. Omissions are first-tier
changes, and this makes it easy to imagine getting from ʿUt ̠mān to C-1. On
the other hand, if C-1 is older, these pluses would be additions of ʿUt ̠mān.
But additions cannot be first-tier changes unless they are candidates for auto-
contamination or some other first-tier process. In fact, several pluses of
ʿUt ̠mān cannot be due to auto-contamination, as they are not attested else-
where in similar contexts. Nor does any other first-tier process explain them.
Therefore, it is difficult to imagine getting from C-1 to ʿUt ̠mān through dic-
tation. In sum, some pluses of ʿUt ̠mān can be omissions of C-1 but not auto-
contaminations of ʿUt ̠mān, a situation consistent with the priority of
ʿUt ̠mān.
The conclusion is that ʿUt ̠mān has the older wording. The pluses of C-1
are due to assimilation of parallels by C-1, while the pluses of ʿUt ̠mān are
omissions by C-1. Some of these omissions of C-1 may have involved (and
been facilitated by) assimilation of parallels.
(4) Assimilation of Nearby Terms. One’s memory is also shaped by the things
one has heard most recently or, if one is familiar with a text, by what one has
previously heard in the vicinity of the point at hand. For this reason, it is
necessary to examine nearby passages to check for the possibility of influences
from nearby terms and parallels. In three instances, ʿUt ̠mān’s phrasing (where
it differs from C-1) parallels nearby passages: Kor 2, 217 (ʿan dīnikum), Kor
2, 221 (bi-id̠nihi), Kor 5, 42 (fa-in ǧāʾūka). On the other hand, there are five
cases in which the variants in C-1 parallel nearby passages: Kor 5, 44
(yaḥ kumūna bi-mā nazzala llāhu fīhā and yaḥ kumu bihi), 5, 45 (ʿalā banī
Isrāʾīla), 63, 3 (d̠ālika bi-annahum qawmun), possibly Kor 63, 11 (li-ġadin),
possibly Kor 5, 45 (li-l-muttaqīna vs. li-qawmin yūqinūna, provided the last
word is yūqinūna and not yuʾminūna).
The greater number of nearby parallels in the case of C-1 might suggest
that C-1 is secondary. However, caution is in order, as the tally of three

against five does not represent a large disparity, and it could be due to chance.
Clearly, it would be desirable to study the other extant folios of the manu-
script to determine whether a trend emerges.
(5) Substitutions. There are eight cases in which each codex uses a different
major word or phrase that cannot be classified under phonetic conservation.
These “substitutions” are listed in Appendix 2, section (c). In these cases, the
two codices generally use synonyms or similar terms both of which appear
elsewhere in similar contexts, or both of which are common terms; however,
there are a couple of exceptions. In the case of Kor 5, 46 (li-l-muttaqīna vs.
li-qawmin yuʾminūna or yūqinūna), it is the version of ʿUt ̠mān that has signif-
icant parallels elsewhere. Normally, this would militate against ʿUt ̠mān’s pri-
ority. (However, if one reads C-1’s variant as yūqinūna instead of yuʾminūna,
then it becomes a candidate for assimilation of a nearby term, consistent with
ʿUt ̠mān’s priority ). On the other hand, the case of Kor 63, 4 (raʾaytahum vs.
ǧāʾūka) leans in the direction of ʿUt ̠mān’s priority. But, there is no trend.
Now, one may also consider the cases in which two different major terms that
sound similar are used. These are included under the rubric “phonetic con-
servation” in Appendix 1, section (b). These do not evince a trend.111
The analysis is complete. The reader will note that I did not call any of
the variants a deliberate change. With a little bit of imagination, anyone
might concoct a theological or political motive for any major variant. (Given
enough zeal for invoking religiously or politically-motivated changes, one can
posit such motives even for variants that turn out to be imaginary.)112 Yet
such explanations are gratuitous in the case at hand, since a single principle
gives a natural explanation for the data: everything is accounted for in a plau-
sible way as a first-tier by-product of semi-orality. Given a falsifiable theory
that explains the data in a simple way, it will be a violation of explanatory
unification and simplicity to use that theory plus another one to explain the
data.113 The hypothesis of deliberate change, unlike the theory of first-tier
change, fails to explain many things. Take, for example, the pluses of C-1.
Say, for the sake of argument, that the scribe of ʿUt ̠mān’s codex deliberately
removed the word “all” in Kor 2, 193 because he did not like it for some
theological or political reason. If so, why would he let “all” remain when the
exact same sentence occurs in Kor 8, 39? Deliberate change fails to explain
111
Another consideration, besides substitutions, that does not reveal a trend is a word’s fre-
quency in the Qurʾān. In both instances, it would be of interest to see whether a trend emerges
in the larger sample of all extant folios of the manuscript.
112
See above p. 363 and footnote 32.
113
On explanatory unification, see Philip Kitcher, “Explanatory Unification”, Philosophy of
Science, 48/4 (1981), p. 507-31.

why C-1 does not add any major element that is not attested elsewhere.Delib-
erate change fails to explain why there are so many minor variations. It does
not explain the use of synonyms. It also fails to explain why most substitutions
obey phonetic conservation.
Lest the appeal to explanatory unification be an overly philosophical way
to put the matter for some readers, it is useful to quote P. Kyle McCarter’s
commonsense formulation of the point:114
Look first for the unconscious error. The better solution to a textual problem—
when one solution assumes a textual accident and another an intentional
alteration—is the first . . . Textual criticism is a rational activity, and there is a
temptation to rationalize the changes that occurred in the text. The fact is, howe-
ver, that deliberate alterations of the text were rare in antiquity, and textual acci-
dents were common. The critic who thinks in terms of inadvertent errors,
therefore, will do the more effective job of restoring the text.
In conclusion, there are several significant cases of asymmetry between the

codices. These disparities are precisely those one would expect to find if the textual
differences were due to the aural dimension of taking dictation and if ʿUt̠mān’s
text were earlier than C-1 in the sense of being closer to the original (or being the
original). This conclusion is falsifiable in principle, meaning that the method
employed here would have objectively led us to the opposite conclusion if the
trends had been the opposite of those observed. The method is not loose
enough to lend itself to justifying any conclusion one desires. This fact has
the following direct practical consequence. Most of the extant folios remain
as yet unanalyzed. In them, one can expect to find exceptions to the trends
identified here. If these exceptions are numerous enough to destroy the
trends, then my conclusions will be undermined. If they are numerous
enough to reverse the trends, then my conclusions will be refuted and
reversed. If, however, the trends remain, then that will be an even more strik-
ing corroboration of ʿUt ̠mānic priority.115
114
Peter Kyle McCarter, Textual Criticism: Recovering the Text of the Hebrew Bible, Philadel-
phia, Fortress Press, 1986, p. 23.
115
To note a caveat, one should also be prepared for the possibility that different sūras have
different transmission histories due to the involvement of different scribes. It could be that a
Companion who collected the Qurʾān did not hear every single sūra directly from the Prophet.
It could be that he heard some sūras personally, but relied on other Companions for the sūras he
had not heard himself. This means different sūras in C-1 could have different transmission his-
tories before they were collected in one codex. The possibility (as maintained in Muslim tradi-
tions) that the Prophet himself disseminated different versions of the sūras or (as indicated in
the Qurʾān) that he revised the sūras, if real, could be expected to make the data more complex.
It may become necessary to consider these potential complications after the remaining folios
have been studied.

The analysis conducted in this section captures a pattern of asymmetry in

the data that shows that the two text types have different characters. This
asymmetry yields a broad conclusion about the relative priority of one text
type over the other. Such an analysis of trends or tendencies, however useful
it may be for arriving at a conclusion about a codex as a whole, is generally
not suitable for judging an individual word or phrase, just as, from the statis-
tical fact that on average one year is warmer than another one, cannot infer
anything with high probability about individual days.
Which Text Type is Older? A Third Method? Obvious Mistakes.
Devin Stewart has made a preliminary attempt to lay out the logic of textual
emendation in the context of copyists’ errors.116 The method, for him, turns
on probability—a probable word that is graphically similar to the given,
improbable text is a candidate for being the original word—and probability
may be assessed, among other ways, by frequency of occurrence: a word that
is well-attested in the Qurʾān is more probable than one that is not.117 He
thus models textual corruption as a movement from a probable word to an
improbable one.
The validity of Stewart’s model for its intended context of written trans-
mission is open to debate; but let us set that question aside and ask whether
the model is applicable to a context for which he did not intend it: namely,
scenarios in which memory plays a more prominent role, as in the case of a
dictated text. In such a situation, it is not obvious that corruption would
tend to run towards greater improbability, as measured, for example, by attes-
tations elsewhere in the book. One would expect that in most cases the oppo-
site would be the case.
Movement toward greater probability would be more common; but it can-
not be denied that Stewart’s model may generate exceptions, and this is one
reason (among others) why the method employed in the last section looks for
statistically significant trends rather than for fixed and inviolable patterns. In
fact, one way to identify candidate instantiations of Stewart’s model is to
focus on the exceptions to the general trend. But then this method is not an
independent way of distinguishing the secondary text type from the primary
116
Devin Stewart, “Notes on Emendations of the Qurʾān”, in The Qurʾān in its Historical
Context, ed. Gabriel S. Reynolds, London, Routledge, 2008, p. 225-48.
117
The link between probability and attestation is shown by Stewart’s comment on Kor 56,
28-31: “It seems odd that bananas are mentioned here, for they do not occur elsewhere in the
Qurʾān” (Ibid., p. 233).

one, for it presupposes that one already has made that determination by iden-
tifying the trend.
Is all hope lost for using Stewart’s approach or a variant thereof? There is a
class of improbable words for which the direction of change is not in doubt,
namely extreme improbabilities that are prima facie errors, and therefore may
be classified as secondary without hesitation. These include obvious errors of
the hand and obvious cases of mishearing or false memory. Both types occur
in C-1. Errors of the hand (see Appendix 2, section (d)) are almost certainly
singular readings that do not represent the text type and, therefore, imply lit-
tle about the relative dates of the two text types; but obvious cases of mis-
hearing or false memory might speak to the degree of accuracy of the codex
as a whole. Instances of the latter in C-1 include mā lahum instead of mā
lahu in Kor 2, 200 and fa-iḫwānuhum instead of fa-iḫwānukum in Kor 2,
220. (Here, I set aside cases of awkwardness in C-1, such as the use of
fa-yaqūlu rabbanā . . . aḫḫirnā . . . fa-aṣsạ ddaqa wa-akun in Kor 63, 10, the use
of the singular al-āya in Kor 2, 219, the use of bihā instead of bihi in Kor 5,
45, and the use of d̠ālika bi-annahum qawmun lā yafqahūna in Kor 63, 3,
which results in the awkward repetition of d̠ālika bi-annahum.)
IV. Conclusion
Implications for the Role of the Prophet
Fidelity in transmission increased rapidly in the early years of Islam. This

proposition is corroborated well when one contrasts the relatively minor dif-
ferences in the ʿUt ̠mānic branch, representing the limited caliber of the
changes that could arise after about AD 650, with the relatively significant
differences between ʿUt ̠mān, C-1, and Ibn Masʿūd. Within the ʿUt ̠mānic tex-
tual tradition, the changes are of the type one would associate with copying
from manuscript or dictating and then correcting against a manuscript. By
contrast, the relatively larger differences with C-1 are due to a measure of
orality. The splitting of the branches took place before the internal develop-
ments in either branch. The branching point represents the prototype, the
state of the text at one remove from the C-1 text type. I will argue that the
prototype should be identified with the Prophet.
Chronology and the Role of the Prophet: The Argument from Stemmatics
The first argument about the identity of the prototype depends on the stem-
mata (p. 394-9). If one takes ʿUt ̠mān as a package (as opposed to a hybrid
based on multiple Companion codices), then among the various stemmata,

(f ) is the most plausible—this is the one in which the prototype P is dictated,

yielding among others ʿUt ̠mān and C-1, which indeed exhibit the kinds of
differences between themselves that would arise with dictation. Now, these
codices go back to the time of the Companions. But the Companions were
close to the Prophet temporally and, certainly in ʿUt ̠mān’s case, personally. The
common ancestor of their codices, the prototype P, would represent an even
earlier stage of the text. But “an earlier stage of the text” when one is speaking
of Companions does not leave much room for maneuver: P would most plau-
sibly be the Prophet’s own recension. As long as one acknowledges that the
codices go back to the era of the Companions, the move to the Prophet as the
prototype is not much of a leap.
The last paragraph proceeded on the basis of stemma (f ). If one rejects
(f ), one shall be left with the less likely scenario (e) in which C-1 derived
from ʿUt ̠mān, and that would push ʿUt ̠mān’s textual tradition back in time
to the period before C-1. But, given C-1’s early date, in this scenario
ʿUt ̠mān’s version is most plausibly identified with the recension of the
Prophet.
Two more scenarios: If one does not wish to acknowledge that there was a
separate prototype, which is fine, then one must consider the remaining
stemmata while removing P, but then it would still follow that ʿUt ̠mān was
prior to C-1, which essentially makes ʿUt ̠mān itself the prototype. In terms of
chronology, this scenario is no different than the ones just outlined. Finally,
if one considers ʿUt ̠mān a hybrid, the stemma is (g). In this scenario, a num-
ber of Companion codices were compared to restore the wording of P. The
early dates of all these Companion codices leads to the same conclusion as
before.
Chronology and the Role of the Prophet: The Internal Evidence

Polarity analysis of internal evidence (p. 399-405) showed that either ʿUt ̠mān
is a more reliable version of the ancestor it shares with C-1 (if one posits a P),
or C-1 derives from ʿUt ̠mān. In light of the early dates of C-1 and ʿUt ̠mān,
these scenarios again lead us back to the Prophet in a manner similar to the
stemmata-based argument.
Learning about the Prophetic Prototype
The results of textual analysis can be compared with early Muslim reports. To
the extent that they agree, there is independent corroboration. But where
they are incompatible, the reports become questionable. Thus, textual analy-
sis is a tool for evaluating early reports. It is particularly gratifying, however,
that textual analysis can also shed light on matters on which literary sources

are silent or inconclusive. For example, historical reports, while suggestive,

do not provide a strong basis for considering one text type as closer to the
Prophet’s own recension than another. Yet textual analysis enables one to
probe this question. The study of the folios so far suggests that the text type
of ʿUt ̠mān preserves the Prophetic prototype better than C-1.
This conclusion provides a broad characterization of the wording of the
Prophetic prototype; but what can be inferred about its specifics? What one
can say about P depends on which stemma one uses. If it is (g), then the
work of textual analysis has already been performed for us, and it would be
hard do better at recovering P than by simply reading ʿUt ̠mān. If it is (e),
then ʿUt ̠mān is necessarily better than the other Companion codices. The rest
of my discussion concerns the more challenging scenario of (f ).
If the true stemma is (f ), then the method is standard: commonalities in
different branches (i.e. different text types) belong to their common ancestor;
and in cases of disagreement, competing claims are weighed against one
another: a majority vote among the branches establishes the original. The
procedure and its outcome are most straightforward when all the text types
agree. The general agreement of the branches on the order in which the verses
are arranged into sūras means that it was their common ancestor, i.e. the Pro-
phetic prototype, that fixed the internal structure of the sūras and the choice
and order of the verses within them. In addition, the wording of the passages
is generally the same in different codices, and common wording goes back to
the Prophetic prototype. But what about the differences?
Going with majority vote, ʿUt ̠mān’s text wins in the vast majority of cases,
and the Prophetic prototype emerges as almost identical with the codex of
ʿUt ̠mān. That is so because given any part of the text, either all three agree
(which is the typical situation), or two agree against one, and in the latter
case ʿUt ̠mān is usually in the majority. There are a number of exceptions, in
which C-1 and Ibn Masʿūd agree against ʿUt ̠mān (see above, Table 4, p. 361).
But there is a problem with the procedure just described. The “one codex,
one vote” motto may be egalitarian, but it is not just. If one source has been
established as more reliable, then it should be given proportionately greater
weight. Therefore, if at some point C-1 and IM agree against ʿUt ̠mān, it is
not clear that they should override it. One must consider other possibilities,
e.g. the Prophet having given revised versions of the prototype over time as
Muslim tradition maintains (thus yielding two versions), cross-contamination,
or, in certain cases, a concurrence of independent adaptations.118
118
One may try to let reported variants of other Companion codices participate in the vot-
ing, but here one faces two challenges: determining which reports are reliable and, more daunt-
ing, ensuring that a reported variant is not a derivative of those that have already participated in
the voting.

What about the verse divisions, i.e. the locations of the verse separators? Is
it possible that the text was divided into verses for the first time after the
Prophet? Or do the divisions go back to the Prophet? There is reason to answer
the latter question in the affirmative: the verse divisions in the non-ʿUt ̠mānic
C-1 agree with the ʿUt ̠mānic divisions, and therefore must have originated
before either, i.e. with the Prophetic prototype. The application of X-Ray fluo-
rescence imaging to Stanford ’07 brought to light verse divisions not other-
wise visible.119 On that leaf, the locations of verse separators are identical to
ʿUt ̠mānic locations (Iraqī system). Moreover, the location of C-1’s 200th verse
marker in sūra 2 is only two or three verses off compared to the ʿUt ̠mānic loca-
tions. This means C-1 and ʿUt ̠mānic verse divisions are usually the same over
a stretch of some 200 verses. The match suggests a genetic link, i.e. common
ancestor. It makes independent development unlikely since one can imagine
alternative ways of separator placement: two independent attempts at dividing
the verses would probably have led to different results.120
The fact that some ʿUt ̠mānic manuscripts have aberrant separator place-
ments does not alter this conclusion. It is a principle of textual analysis that if
different branches of a textual tradition include manuscripts with shared dis-
tinctive features, then the existence of manuscripts that do not share those
features does not preclude assigning those shared features to the common
ancestor.121 The principle is illustrated in the diagram below. Symbols ‘a’, ‘b’,
and ‘c’ represent a textual feature, say, a distinctive verse separator placement
pattern. One ought to ascribe the feature ‘a’ to the common ancestor (O) even
though it is not found in every manuscript. Because it is distinctive, its pres-
ence in the two independent branches is best described by genealogical descent
from a common ancestor, and its absence in O12 and O22 by corruption.
This conclusion would be suspended only if a good case were made for cross-
contamination or independent convergent adaptations.
The above comments apply exclusively to the contents of the sūras. But
what about the order in which the sūras are arranged relative to one another?
Is it possible to infer the order in the Prophetic prototype? The evidence
includes the sūra orders reported for Ibn Masʿūd and Ubayy b. Kaʿb. These
reports must be taken seriously, in part because they are partially corrobo-
rated by C-1. In an attempt to ascertain the order of their common ancestor,
119
When the attempt is made to study verse division practices in the entire extant portion of
C-1, a caveat should be kept in mind: the fact that X-Ray fluorescence imaging brought into
view verse separators that are not otherwise visible shows that in general if we do not see a verse
separator in the lower text, that does not always mean the absence of one.
120
Verse-separators on Stanford ’07 that are “non-obvious” include those at Kor 2, 191, 198,
200, and 201. There are also spots without a verse separator at which one would have been
imaginable.
121
See, e.g., the rules stated in Reynolds and Wilson, Scribes & Scholars, p. 211-2.

O1 O2
O11 O12 O21 O22

a b a c
one may again conduct a vote. However, voting does not yield a definite
result, since each codex has a different sūra order. The orderings do have com-
monalities, and these might go back to the Prophetic prototype; but the
amount of disagreement is significant. The conclusion, then, is that the
Prophet probably did not fix the order of the sūras, except possibly in some
cases. Such a conclusion is not new. Pre-modern scholars advanced a number
of theories about whether the Prophet had fixed the order of the sūras. The
majority believed he had not.122 Ibn ʿAt ̣iyya (d. 542/1148), however, sug-
gested that the Prophet arranged some of the sūras but not all.123
Implications for the nature of ʿUt̠mān’s contribution
I make a key distinction between two types of evidence: (1) the collective,
consistent, independent memory of the early communities on matters that
we expect the communities to have known about, and (2) the anecdotes of
individuals (especially on matters that the entire community could not be
expected to have known about).124 The former is reliable, while the latter
requires evaluation. As discussed above, based (among other things) on the
collective memory of communities, we know that ʿUt ̠mān provided the cities’
master codices. However, the exact details of how the ʿUt ̠mānic text was
122
Al-Suyūṭī, al-Itqān, p. 171.
123
Ibn ʿAt ̣iyya, al-Muḥ arrar, I, p. 50. My attention was drawn to this source by al-Aʿẓamī, The
History of the Qurʾānic Text, p. 78. For Ibn ʿAt ̣iyya see, e.g., Ḫ ayr al-Dīn al-Ziriklī, al-Aʿlām:
Qāmūs al-tarāǧim li-ašhar al-riǧāl wa-l-nisāʾ min al-ʿarab wa-l-mustaʿribīn wa-l-mustašriqīn, 5th
ed., Beirut, Dār al-ʿilm li-l-malāyīn, 1980, III, p. 282.
124
Cf. Modarressi, “Early Debates”, p. 13-4. For a discussion of why the collective memory
can be considered reliable, see above p. 365-6.

put together was not something the entire community knew, and it was
remembered in different ways. This is not surprising: the people of, say, Kūfa,
as a whole could be expected to remember the name of the caliph who sent
them their master codex, as the event would have been a very public affair,
but relatively few would have had direct knowledge of the pre-history of that
codex. As a result, there are conflicting trends in the traditions about the pro-
cess that resulted in the standard text.
As Modarressi has noted, the very fact that ʿUt ̠mān standardized the
Qurʾān suggests that there were different versions.125 The burning question is
how far-reaching their differences were. Did the codices arrange existing
verses into sūras in different ways, so that two versions of a sūra could have
different verses, or have the same verses but order them differently? Or were
the differences limited to the rather modest kind attested in C-1? More
fundamentally, who arranged existing verses into sūras: was it ʿUt ̠mān’s
team or the Prophet? On this question there was no single collective memory,
but rather divergent understandings handed down to us by individuals.
Two sets of reports of individuals support two very different positions.
One group indicates that it was ʿUt ̠mān’s team who strung existing verses
together to form the sūras, while another group ascribes this act to the
Prophet.126 Among reports that support ascription to the Prophet, one may
include many of those describing the Companion codices: it would be mean-
ingless to say that different Companions ordered the sūras differently if the
sūras had not been formed already.127 Furthermore, al-Aʿmaš’s methodical col-
lation of Ibn Masʿūd’s verbal differences with ʿUt ̠mān and reports about other
Companion codices presuppose that the codices were the same except for the
specified differences at the level of words and phrases. The sequences of sen-
tences and verses within the sūras are the same in the known Companion
codices, except for rare cases of transpositions of short adjacent verses.
Modern scholars, like pre-modern ones, have generally opted for one of
these two positions. For example, Jeffery accepted the ascription to ʿUt ̠mān
and therefore rejected the reports about the different sūra orderings of other
Companions. On the other hand, Alford Welch and Hossein Modarressi
125
Modarressi, “Early Debates”, p. 14.
126
Modarressi (“Early Debates”) provides a brief survey of traditions suggesting that the
sūras were fixed only after the Prophet’s death (p. 8-13) before describing them as unhistorical.
He calls them “extremely problematic” on p. 14 and goes on to propose a hypothesis explaining
how they were generated. For another useful selection of the divergent traditions (though not
a very convincing attempt to harmonize them), see al-Aʿẓamī, The History of the Qurʾānic Text,
p. 71-101.
127
A.T. Welch, R. Paret, J.D. Pearson, “al-Ḳ urʾān”, EI 2.

favor the idea that the sūras were put in their final forms during the Prophet’s
lifetime. The study of C-1 has confirmed the latter position.
Implications for the Historicity of Companion Codices
Manuscript evidence now corroborates pre-modern reports about the exis-

tence of Companion codices, their having different sūra orderings, and, to an
extent, the nature of their verbal differences. Conclusively refuted is John
Burton’s theory that all such reports were post-ʿUt ̠mānic fictions aimed at
“countering, ellucidating, or even evading the ʿUt ̠mān text.”128 But once one
acknowledges the traditional notion of Companion codices, it remains to
explain their differences. Such differences would have arisen as the Prophet
recited the sūras and different scribes wrote them down.
Given that believers were urged to recite the Qurʾān as much as feasible
(Kor 73, 20), the question arises whether the Prophet knew of some of these
differences, and, if he did, what his reaction was. Ibn Masʿūd may or may not
have spoken the words ascribed to him in the following report, but there is
no doubt that they embody one of the earliest theories explaining the differ-
ences among the Companion codices:
Abū Usāma—Zuhayr—al-Walīd b. Qays—ʿUt ̠mān b. Ḥ assān al-ʿĀmirī—Filfila

l-Ǧ uʿfī: he [i.e. this last] said: I was among those who went to ʿAbd Allāh [Ibn
Masʿūd], fearful over [the issue of ] the codices. We came into his presence and a
man in the group said, “We have not come to you as guests; we came here upon
being frightened by this news [of the standardization].” He answered, “The
Qurʾān was sent down to your prophet through seven [heavenly] gates according
to seven modes (aḥ ruf, ḥ urūf ). Previous to you [i.e. before your generation], the
Book was sent down through a single gate according to a single mode (ḥ arf ).129
The report presupposes that the Prophet approved of Ibn Masʿūd’s version of
the Qurʾān alongside that promulgated by ʿUt ̠mān. It is not inconceivable
that different scribes read different versions back to the Prophet, and were
128
John Burton, The Collection of the Qur’an, Cambridge, Cambridge University Press, 1977,
p. 228. Burton assumed that if one can imagine a theological motive for creating a report, the
report cannot be historically true. The problem is that often with a little imagination one can
concoct theological motives even for accurate reports.
129
Ibn Abī Dāwūd, al-Maṣāḥ if, p. 18. Edward Lane, when defining ḥ arf in his Lexicon, cites
this expression: fulān yaqraʾu bi-ḥ arf Ibn Masʿūd. From the report quoted by Ibn Abī Dāwūd and
other evidence, it appears that the “modes” (aḥ ruf ) in which the Qurʾān is said to have been
revealed originally encompassed the Companion codices. This also happens to have been the
position of most premodern and early-modern scholars. See above, p. 346 and footnote 4.

met with his tacit approval when they did so.130 Yet, it must be stressed that at
present there is no definitive evidence for or against this. In any case, if the
Prophet did tacitly approve more than one version, that would not necessar-
ily mean that all versions represented equally precisely the recitations as they
left his mouth.
Summary
Muḥammad dictated the revelations, and scribes wrote them down. This gave
rise to a number of Companion codices. As the Prophet had not fully deter-
mined the order of the sūras relative to one another, these codices had differ-
ent sūra orders. However, he had fixed the contents of the sūras, including the
distribution of verses within them and even the verse divisions. On these ele-
ments, and especially where the actual text is concerned, the codices showed
great agreement.
Yet, the aural dimension of the Prophet’s dictation at times generated
changes, giving rise to occasional verbal differences. Many of the differences
among the Companion codices point to semi-orality, and they go back to the
Companions’ transcription of a Qurʾān recited by the Prophet. If the scribes
recited the text back to the Prophet—and we do not know whether this hap-
pened—one wonders if the Prophet tacitly endorsed some of these differ-
ences, relatively small as they generally seem to be. If so, that would not
negate the fact that one version better represented what the Prophet himself
actually recited; but which one?
The caliph ʿUt ̠mān established the standard version, an undertaking that,
according to the literary sources, involved a committee, and, above all, a
scribe of the Prophet named Zayd b. T̠ābit.131 If ʿUt ̠mān formed a committee
to deal with this potentially explosive issue, then that was a politically astute
move, making it easier to gain the acceptance of a large part of the commu-
nity, and helping deflect criticism from the caliph himself.132 And if it is true
130
This scenario fits a perhaps improbable reading of Kor 25, 5. The verse runs: wa-qālū
asāṭīru l-awwalīna ktatabahā fa-hiya tumlā ʿalayhi bukratan wa-aṣīlā. It says of Muḥammad’s
opponents in Mecca that, “They say, ‘[that the Qurʾān is] fairy-tales of the ancients that
he [Muḥammad] has caused to be written down, so that they are recited (or dictated) to him
at the dawn and in the evening.’” This translation closely follows those of Arberry, Asad,
Sher Ali, and Abdalhaqq and Aisha Bewley. The improbable reading would be that the
Prophet would have followers write the Qurʾān, by dictation, and it would be recited back to
him (tumlā ʿalayhi). The perhaps more probable reading is that enemies charged that informants
recited the tales to the Prophet day and night. Either way, the implication is that the Qurʾān
was written.
131
Rāmyār, Tārīḫ-i Qurʾān, p. 416-35; A.T. Welch, et al., “al-Ḳ urʾān”.
132
Michael Cook, “Concluding Remarks”, Colloquium on the Evidence for the Early

that the outcome was a hybrid codex, then that had the added political advan-
tage that Companions would not be chagrined to learn that a fellow Compan-
ion’s codex was preferred over theirs.133
In any case, textual criticism suggests that the standard version is the most
faithful representation, among the known codices, of the Qurʾān as recited by
the Prophet. This appears, at first, as a curious coincidence; but on second
thought it is not surprising: if anybody had the resources to ensure that a reli-
able version be chosen, it would have been the caliph; and if anybody had
more to lose by botching up the task, again that would have been ʿUt ̠mān,
whose political legitimacy and efficacy as caliph depended completely on the
good will of fellow distinguished associates of the Prophet. The remarkably
few and minor skeletal-morphemic differences among the codices ʿUt ̠mān
sent to the cities is another indication of the care that was put into the pro-
cess of standardization.134
Not everybody considered the standard version as the only legitimate one.
Some, espousing a sort of codical pluralism, continued to consider other
Companion codices as legitimate. The codex of Ibn Masʿūd, in particular,
continued to have supporters in the first two centuries AH. Nonetheless,
upon its dissemination the standard version quickly became predominant
everywhere. Given the vast expanse and decentralized nature of the empire,
the center’s intervention had achieved a remarkable degree of success. Several
years after Medina’s act of standardization, however, a new era was ushered
in. The murder of ʿUt ̠mān in AD 656 occasioned what came to be known as
the “First Civil War” in the historical memory of Muslims. It polarized and
fractured the community irrevocably. Had empire-wide standardization been
attempted at any moment after this point, it would have been a hopeless
undertaking.
Post-Script: Directions for Future Research
It is a curious and telling fact of the field of Qurʾānic studies that few of its
achievements in the West have been celebrated more than a project that never
gathered much steam to begin with, namely the plan by Bergsträsser and his
History of the Qurʾān, Stanford University, July 31, 2009. Rāmyār, too, points out that Ibn
Masʿūd’s attacks were directed against Zayd (Tārīḫ-i Qurʾān, p. 418).
133
Rāmyār, Tārīḫ-i Qurʾān, p. 422.
134
The variants typically affect one or two characters without changing the meaning, and
they number thirty-six. As Cook notes, “This is not a large number for a text as long as the
Koran; the proportion of words affected is less than one in two thousand. The number of dis-
tinct readings at any given point never exceeds two, each regional codex having one or the
other” (Cook, “The Stemma of the Regional Codices of the Koran”, p. 90-1).

collaborators to prepare a critical text of the scripture.135 And it is perhaps in

keeping with this unreal dimension that now we learn that the oft-repeated
reason for its termination was a deliberate lie: the tale was that Allied bomb-
ing during WWII destroyed a collection of photographs of early qirāʾāt works
and photographs of Qurʾāns in the Kūfī script. Yet in the midst of the elegies
in the past decades that put the blame on bombs, few seemed to notice that
early Qurʾān manuscripts that have been available and accessible in Europe
and the Muslim world were not being studied.136 In fact, only in the last few
years have some of them been subjected to textual analysis by scholars such
as Yasin Dutton. Nor was it apparent to all that if the study of qirāʾāt moved
at a glacial pace, this had nothing to do with lack of materials.
It is obvious that the study of the early history of the Qurʾān requires a
dual track involving (1) the qirāʾāt and early traditions, and (2) manuscripts.
This two-pronged approach was, for example, what Arthur Jeffery saw over
eighty years ago as the basis of his collaboration in the above-mentioned proj-
ect. A preliminary step would have been to collect the “textual variants gath-
ered from the Commentaries, Lexica, Qirāʾāt books and such sources”,137 a
task that would be “superhuman” if done in a comprehensive manner.138
The dream that one day all the qirāʾāt variants would be collected has been
fulfilled, although the fulfillment has not received, in the West, anything like
the recognition accorded to Bergsträsser’s and Jeffery’s dream. Many qirāʾāt
works have been published and some have been analyzed in the Muslim
world. There are now also encyclopedias: the Muʿǧam al-qirāʾāt al-qurʾāniyya
of Aḥmad Muḫtār ʿUmar and ʿAbd al-ʿĀl Sālim Mukarram has gone through
two editions in Kuwait and at least one printing in Iran.139 In addition, ʿAbd
al-Lat ̣īf Muḥammad Ḫ aṭīb’s comprehensive Muʿǧam al-qirāʾāt was published
in Syria. It is an erudite work that canvasses works of exegesis, grammar, and
qirāʾāt. These efforts have mapped the literature and paved the way for ana-
lytical studies. Building on them is a desideratum of the field of Qurʾānic
studies.
As for the study of manuscripts, it may appear to some that what the field
needs most is a database of images of all Qurʾān manuscripts. That would
undoubtedly be an invaluable resource, but let us be clear about what it could
135
For a summary of the critical text project, see Jeffery, Materials, p. vii.
136
It is the lack of systematic textual studies of manuscripts to which I am referring. There
have been useful paleographic studies by Abbott, Munajjid, Déroche, Whelan, and others. See
the sources cited in the works listed in footnote 50.
137
Jeffery, Materials, p. vii.
138
Jeffery, as quoted in Rabb, “Non-Canonical Readings”, p. 87.
139
Aḥmad Muḫtār ʿUmar, ʿAbd al-ʿĀl Sālim Mukarram, Muʿǧam al-qirāʾāt al-qurʾāniyya:
maʿa muqaddima fī l-qirāʾāt wa-ašhar al-qurrāʾ, 2nd edition, Kuwait, D̠āt al-Salāsil, 1988.

or could not help achieve. Obviously, non-ʿUt ̠mānic manuscripts would tell
us much about the early history of the Qurʾān, but so far there is only one
known manuscript in this category, namely Ṣanʿāʾ 1. When one speaks of col-
lecting all manuscripts, one means essentially the innumerable ʿUt ̠mānic
materials. The main aim of this gargantuan task would be to reconstruct their
archetype, the original codex of ʿUt ̠mān. But as explained above (p. 368), the
work of Michael Cook and others has shown that we already know the skele-
tal-morphemic text of the original codices of ʿUt ̠mān, meaning that the
uncertainty of the skeletal text primarily concerns spelling. A list of all the
ways in which the words in ʿUt ̠mānic manuscripts differ from that known
text would be nothing more than a list of later accretions, useful at best for
tracing the history of manuscript production after AD 656. New information
might clarify certain spelling practices, some disputed undotted consonants
and disputed verse divisions in the original codices sent out by ʿUt ̠mān; but
while any new tidbit would be exciting for the Qurʾān specialist, rarely will
such information make a major difference in our understanding of the his-
tory of the Qurʾān.
The amount of work yet to be done is great, and the main paths of
embarking on the tasks are clear. It is now equally clear that recent works in
the genre of historical fiction are of no help. By “historical fiction” I am refer-
ring to the work of authors who, contentedly ensconced next to the moun-
tain of material in the premodern Muslim primary and secondary literature
bearing on Islamic origins, say that there are no heights to scale, nothing to
learn from the literature, and who speak of the paucity of evidence. Liberated
from the requirement to analyze the literature critically, they can dream up
imaginative historical narratives rooted in meager cherry-picked or irrelevant
evidence, or in some cases no evidence at all. They write off the mountain as
the illusory product of religious dogma or of empire-wide conspiracies or
mass amnesia or deception, not realizing that literary sources need not always
be taken at face value to prove a point; or they simply pass over the mass of
the evidence in silence. A pioneering early example of such historical fiction
was Hagarism, written by Patricia Crone and Michael Cook. While few spe-
cialists have accepted its narrative, the book has nevertheless profoundly
shaped the outlook of scholars. It has given rise to a class of students and edu-
cators who will tell you not only that we do not know anything about Islamic
origins, but also that we cannot learn anything about it from the literary
sources. All this would be good and well if the mountain of evidence had
been studied critically before being dismissed as a mole hill; but the modern
critical reevaluation of the literary evidence has barely begun. And, signifi-
cantly, any number of results have already demonstrated that if only one

takes the trouble to do the work, positive results are forthcoming, and that
the landscape of the literary evidence, far from being one of randomly-scat-
tered debris, in fact often coheres in remarkable ways. A good example of
such findings would be some of Michael Cook’s own fruitful recent studies in
the literary sources in two essays of his already discussed here. It is not his
confirmation of some elements of the traditional account of the standard
Qurʾān that I wish to highlight here, noteworthy as it may be, but rather his
demonstration that we can learn from the study of the literary sources.
Appendix 1. Lower Text: (a) Minor Elements and (b) Phonetic Conserva-
tion of Major Elements
Items placed in parentheses (x) are only partly visible, but enough is visible to
have good reason for the readings given here.
Items placed in double parentheses ((x)) are not visible, and the readings
offered here are usually largely speculative.
(a) Minor Elements
For the definition of “minor”, see above, p. 386.
Verse ʿUt ̠mān C-1 Type of Variant

2, 200 ‫ﻣﺎ ﻟﻪ‬ ‫ﻣﺎ ﻟﻬﻢ‬ minor (person)
Note: the C-1 version is problematic. The personal pronoun refers to man,
and it should be singular.
2, 210 ‫ﻳﻨﻈﺮﻭﻥ ﺇﻻ ﺃﻥ ﻳﺄﺗﻴﻬﻢ‬ ‫ﺗﻨﻈﺮﻭﻥ ﺇﻻ ﺃﻥ ﻳﺄﺗﻴﻜﻢ‬ minor (person)

2, 220 ‫ﻓﺈﺧﻮﺍﻧﻜﻢ‬ ‫ﻓﺈﺧﻮﺍﻧﻬﻢ‬ minor (person)
Note: the C-1 version makes little sense, while the version of ʿUt ̠mān reads
naturally.
5, 45 ‫ﻟﻪ‬ (‫ )ﻟﻜﻢ‬minor (person)
5, 51 ‫ﺑﻌﻀﻬﻢ‬ ‫ ﺑﻌﻀﻜﻢ‬minor (person)
63, 1 ‫ﻧﺸﻬﺪ‬ (‫ ﻳﺸﻬﺪﻭ) ﺍ‬minor (person)
63, 4 ‫ﻓﺎﺣﺬﺭﻫﻢ‬ ((‫ ))ﻓﺎﺣﺬﺭﻭﻫﻢ‬minor (person)

(cont.)
63, 10 ‫ ﻓﻴﻘﻮﻝ ﺭﺏ‬. . . ‫ ﺃﺧﺮﻧﺎ‬... ‫ﻓﻴﻘﻮﻝ ﺭﺑﻨﺎ‬ minor (person)

‫ ﻓﺄﺻﺪﻕ‬. . . ‫ﺃﺧﺮﺗﻨﻲ‬ ‫ﻓﺄﺻﺪﻕ‬
‫ﻭﺃﻛﻦ‬ (‫)ﻭﺃﻛﻦ‬
62, 3 ‫ ﺑﻬﻢ‬. . . ‫ﻣﻨﻬﻢ‬ ( ‫ ﺑﻬﻢ‬. . . ‫)ﻣﻨﻜﻢ‬ minor (person)

2, 191 ‫ﻳﻘﺘﻠﻮﻛﻢ ﻓﻴﻪ‬ ‫ﻳﻘﺘﻠﻮﻛﻢ‬ minor omission
‫ﻓﺈﻥ ﺃﺣﺼﺮﺗﻢ‬ ‫ﺇﻥ ﺃﺣﺼﺮﺗﻢ‬

( fīhi )
2, 196 minor omission ( fa)
2, 196 ‫ﻓﻔﺪﻳﺔ‬ ‫ﻓﺪﻳﺔ‬ minor omission ( fa)
Apparent error of the
‫ﻣﻦ ﻗﺒﻠﻪ‬ ‫ﻗﺒﻠﻪ‬

hand
2, 198 minor omission
‫ﺳﺒﻴﻞ ﺍﷲ‬ (‫)ﺳﺒﻴﻠﻪ‬

(min)
‫ﺃﻋﺮﺽ ﻋﻨﻬﻢ‬ ‫ﺃﻋﺮﺽ‬

(allāh)
‫ﺇﻧﺎ ﺃﻧﺰﻟﻨﺎ‬ ‫ﻭﺃﻧﺰﻟﻨﺎ‬

(ʿanhum)
‫ﻭﺃﺣﺬﺭﻫﻢ ﺃﻥ‬ (‫)ﻭﺃﺣﺬﺭ ﺃﻥ‬

(innā)
‫ﺍﻧﺎ ﻟﻤﻨﺠﻮﻫﻢ‬ ‫ﺍﻧﺎ ﻣﻨﺠﻮﻫﻢ‬

(hum)
‫ﻟﺮﺳﻮﻟﻪ‬ ((‫))ﺭﺳﻮﻟﻪ‬
15, 59 minor omission (la)
‫ﻣﻦ ﺍﻟﻠﻬﻮ ﻭﻣﻦ ﺍﻟﺘﺠﺮﺓ‬ ‫ﻣﻦ ﺍﻟﻠﻬﻮ ﻭﺍﻟﺘﺠﺮﺓ‬

63, 1 minor omission (la)
‫ﻓﻴﻮﻣﺌﺬ‬ ((‫))ﻳﻮﻣﺌﺬ‬
(wa-min)
89, 25 minor omission ( fa)
2, 194 ‫ﺑﻤﺜﻞ ﻣﺎ ﺍﻋﺘﺪﻯ ﻋﻠﻴﻜﻢ ﺑﻪ ﺑﻤﺜﻞ ﻣﺎ ﺍﻋﺘﺪﻯ ﻋﻠﻴﻜﻢ‬ minor addition (bihi)
2, 217 ‫ﻗﺘﺎﻝ ﻓﻴﻪ‬ ‫ﻭ)ﻋﻦ( ﻗﺘﻞ ﻓﻴﻪ‬ minor addition
(wa-ʿan)
2, 197 ‫ﻓﻼ ﺭﻓﺚ‬ (( ‫ﻓﻼ ﺭﻓﺚ ))ﻓﻴﻬﻦ‬ minor addition
(fīhinna)
C-1’s version is a candidate for the assimilation of a nearby term.

(cont.)
2, 204 ‫ﰲ ﺍﻟﺪﻧﻴﺎ‬ ‫ﰲ ))ﻫﺬﻩ(( ﺍﻟﺪﻧﻴﺎ‬ minor addition

(hād̠ihi)
C-1’s version is a candidate for the assimilation of parallels at Kor 3, 117; 7,
156; 11, 60; 16, 30; 20, 72; 28, 42; 29, 64; 39, 10; and 40, 39.
2, 194 ‫ﻓﻤﻦ‬ ‫ﻭﻣﻦ‬ minor substitution
‫ﺑﺎﻟﻌﻤﺮﺓ‬ ‫ﺑﻌﻤﺮﺗﻪ‬
( fa vs. wa)
2, 196 minor substitution
(possessive vs.
‫ﻓﻀﻼ‬ ‫ﺍﻟﻔﻀﻞ‬
definite)
2., 98 minor substitution
(definite vs.
‫ﻓﺈﺫﺍ‬ ‫ﻭﺇﺫﺍ‬
indefinite)
‫ﻓﻠﻦ ﯾﻀﺮﻭک‬ ‫ﻻ ﯾﻀﺮﻭک‬

( fa vs. wa)
(lan vs. lā)
For a similar substitution, but in the opposite direction, see below, Kor 62, 7.
5, 43 ‫ﻭﻣﺎ ﺍﻭﻟﺌﻚ‬ ‫ﻓﻤﺎ ﺍﻭﻟﺌﻚ‬ minor substitution
‫ﻟﻠﺬﻳﻦ‬ ‫ﻭﺍﻟﺬﻳﻦ‬
(wa vs. fa)
‫ﻓﺴﻮﻑ‬ ‫ﻓﺴ‬
(li vs. wa)
(synonyms sawfa vs.
‫ﻭﻟﺮﺳﻮﻟﻪ ﻭﻟﻠﻤﺆﻣﻨﻴﻦ‬ (‫)ﻭﻟﻠﺮﺳﻮﻝ ﻭﺍﻟﻤﺆﻣﻨﻴﻦ‬

sa)
(possessive vs.
‫ﻭﻻ ﻳﺘﻤﻨﻮﻧﻪ‬ ‫ﻭﻟﻦ ﻳﺘﻤﻨﻮﻩ‬

definite)
(lā vs. lan)
For a similar substitution, but in the opposite direction, see above, Kor 5, 42.

(cont.)
(b) Phonetic Conservation of Major Elements
For the definition of “major”, see above, p. 387.

2, 196 ‫ﺍﺳﺘﻴﺴﺮ‬ ‫ﺗﻴﺴﺮ‬ phonetic
conservation
‫ﻓﻤﺎ ﺍﺳﺘﻴﺴﺮ‬ ‫ﻓﻤﺎ ﺗﻴﺴﺮ‬

(same-root synonym)
2, 196 phonetic
conservation
‫ﺍﻷﻳﺖ‬ ((‫))ﺍﻷﻳﺔ‬
(same-root synonym)
2, 219 phonetic
conservation
(singular vs. plural)
The plural form (in ʿUt ̠mān) reads more naturally.
Elsewhere in the Qurʾān, the phrase yubayyinu llāhu is always followed by the
plural form āyāt. See Kor 2, 187; 2, 219; 2, 242; 2, 266; 3, 103; 5, 89; 24, 18;
24, 58; 24, 59; 24, 61.
2, 222 ‫ﻓﺈﺫﺍ ﺗﻄﻬﺮﻥ‬ (‫ﻓﺈﺫﺍ) ﻃﻬﺮﻥ‬ phonetic
conservation
‫ﻭﻟﻬﻢ ﻓﻲ ﺍﻻﺧﺮﺓ‬ ‫ﻭ)ﻓﻲ( )ﺍﻻﺧﺮ(ﺓ ﻟﻬﻢ‬

(same-root verb)
5, 41 phonetic
conservation (word
‫ ﺃﻛﻠﻮﻥ‬. . . ‫ﺳﻤﻌﻮﻥ‬ ‫ ﺃﻛﻠﻮﻥ‬. . . (‫)ﻳﺴﻤﻌﻮﻥ‬

order: transposition)
5, 42 phonetic
conservation
(same-root alternative)
5, 43 ‫ﻭﻛﻴﻒ ﻳﺤﻜﻤﻮﻧﻚ‬ ‫ﻭﻛﻴﻒ ﻳﺤﻜﻤﻮﻧﻚ‬ phonetic
‫ﻭﻋﻨﺪﻫﻢ ﺍﻟﺘﻮﺭﻳﺔ ﻓﻴﻬﺎ‬ ‫ﻭﺍﻟﺘﻮﺭﻳﺔ ﻓﻴﻬﺎ ﺣﻜﻢ ﺍﷲ‬ conservation (word
‫ﺣﻜﻢ ﺍﷲ‬ (‫)ﻋﻨﺪﻫﻢ‬
5, 43 ‫ﻫﺪﻯ ﻭﻧﻮﺭ‬ (‫ﻧﻮﺭ ﻭ)ﻫﺪﻯ‬ phonetic

conservation (word

(cont.)
5, 43 ‫ﺍﻟﻨﺒﻴﻮﻥ‬ ‫ﺍﻷﻧﺒﻴﺎﺀ‬ phonetic
conservation
(equivalent plural
forms)
‫ﻣﻦ ﺗﺼﺪﻕ ﺑﻪ‬ ‫)ﻓﺈﻥ( ﺗﺼﺪﻗﻮﺍ ﺑﻬﺎ‬

Both forms appear in the standard Qurʾān.
5, 45 phonetic
conservation
(syntactical structure
plus minor elements,
person and gender)
In C-1, due to illegibility, I cannot tell if what follows is fa-huwa or fa-hiya.
ʿUt ̠mān’s term bihi seems more appropriate than C-1’s term bihā.
For another instance of C-1’s propensity for constructing conditionals with

in instead of man, see below, Appendix II (c), Kor 2, 196. Both structures are
well-attested in the Qurʾān.
5, 47 ‫ﺃﻧﺰﻝ ﺍﷲ ﻓﻴﻪ‬ ‫ﺃﻧﺰﻝ )ﻋﻠﻴﻬﻢ( ﻓﻴﻪ‬ phonetic
conservation (active
vs. passive, plus
different minor
elements)
The passive form unzila followed by the preposition ʿalā occurs eleven times
(Kor 2, 91; 3, 84; 6, 8; 6, 157; 10, 20; 11, 12; 13, 7; 13, 27; 25, 21; 29, 50;
38, 8).
The active form anzala occurs some sixty-six times, including twenty-nine
times followed immediately by llāh.
‫ﻟﻜﻞ ﺟﻌﻠﻨﺎ ﻣﻨﻜﻢ ﺷﺮﻋﺔ‬ ‫ﻟﻜﻞ ))ﻣﻨﻜﻢ(( ﺟﻌﻠﻨﺎ‬
(‫)ﺷﺮﻋﺔ‬
5, 48 phonetic
conservation (word
‫ﻣﻦ ﺍﷲ ﺣﻜﻤﺎ‬ ‫)ﺣﻜﻤﺎ( ﻣﻦ ﺍﷲ‬

or minor omission
5, 50 phonetic
conservation (word

(cont.)

5, 52 ‫ﻓﻴﺼﺒﺤﻮﺍ‬ (‫ﻓﻴﺼﺒﺤﻮ)ﻥ‬ phonetic
conservation (verb in
‫ﺗﺆﻣﺮﻭﻥ‬ ‫ﻳﺎﻣﺮﻭﻥ‬
different aspect)
15, 65 phonetic
conservation (active
vs. passive, plus
person)
63, 1 ‫ﻭﺍﷲ ﻳﺸﻬﺪ‬ (‫ﻭﻳﺸﻬﺪ )ﺍﷲ‬ phonetic
conservation (word
‫ﻳﺘﻤﻨﻮﻧﻪ‬ ‫ﻳﺘﻤﻨﻮﻩ‬
62, 7 phonetic
conservation
(yatamannawnahu
vs. yatamannawhu)
As for ʿUt ̠mān’s version, the word yatamannawnahu is found nowhere else in
the Qurʾān, although in Kor 3, 143 we have kuntum tamannawna l-mawta.
As for C-1, the entire verse here is found also in Kor 2, 195 with C-1’s wording,
yatamannawhu.
Appendix 2. Lower Text, Major Elements: (a) Omissions, (b) Additions,

(c) Substitutions, (d) Scribal errors, (e) Illegible variants
Items placed in double parentheses ((x)) are not visible, and the readings
offered here are usually largely speculative.
(a) “Omissions” in C-1 (i.e. pluses of ʿUt ̠mān)

(C-1 relative to
ʿUt ̠mān)
2, 196 ‫ﻭﻻ ﺗﺤﻠﻘﻮﺍ ﺭﺅﺳﻜﻢ‬ ‫ﻭﻻ ﺗﺤﻠﻘﻮﺍ‬ omission (ruʾūsakum)

(cont.)
(C-1 relative to
ʿUt ̠mān)
The version in ʿUt ̠mān has something of a parallel elsewhere: the word
ruʾūsakum and the verb ḥ alaqa occur only in one other spot in the Qurʾān, at
Kor 48.27, in muḥ alliqīna ruʾūsakum.
‫ﻣﻦ ﺻﻴﺎﻡ ﺃﻭ ﺻﺪﻗﺔ ﺃﻭ‬ ‫ﻣﻦ ﺻﻴﻢ ﺃﻭ ﻧﺴﻚ‬
‫ﻧﺴﻚ‬
2, 196 omission in a list (aw
ṣadaqatin)
ʿUt ̠mān has min ṣiyāmin aw ṣadaqatin aw nusukin, whereas C-1 gives min
ṣiyāmin aw nusukin.
Neither version has parallels elsewhere in the Qurʾān.
The omitted word in C-1, ṣadaqa, occurs six times in ʿUt ̠mān (Kor 2, 196; 2,
263; 4, 114; 9, 103; 12, 26; 58, 12). Its plural, ṣadaqāt, occurs eight times (2,
264; 2, 271; 2, 276; 9, 58; 9, 60; 9, 79; 9, 104; 58, 13).
2, 201 ‫ﻓﻲ ﺍﻟﺪﻧﻴﺎ ﺣﺴﻨﺔ ﻭﻓﻲ‬ ‫ﻓﻲ ﺍﻟﺪﻧﻴﺎ ﻭﺍﻷﺧﺮﺓ ﻭﻗﻨﺎ‬ omission (ḥ asana . . .
‫ﺍﻷﺧﺮﺓ ﺣﺴﻨﺔ ﻭﻗﻨﺎ‬ fī . . . ḥ asana)
The closest parallel is Kor 7, 156: wa-ktub lanā fī hād̠ihi l-dunyā ḥ asanatan
wa-fī l-āḫirati; cf. 16, 30; 16, 41 and 16, 122. The phrase “this world and the
next” (l-dunyā wa-l-āḫira) occurs fifteen other times in the Qurʾān. The word
ḥ asana occurs twenty-six additional times in the Qurʾān.
Neither the version in ʿUt ̠mān nor that in C-1 can be considered unexpected,
and both make sense in context.
2, 213 ‫ﺑﻐﻴﺎ ﺑﻴﻨﻬﻢ‬ omission (baġ yan
baynahum)
ʿUt ̠mān’s phrase baġ yan baynahum appears in three other spots in similar
contexts: Kor 3, 19; 42, 14; 45, 17. The term baġ yan is also used alone in
2, 90 and 10, 90.
2, 213 ‫ﺍﻟﺬﻳﻦ ﺧﻠﻮﺍ ﻣﻦ ﻗﺒﻠﻜﻢ‬ ‫ﺍﻟﺬﻳﻦ ﻣﻦ ﻗﺒﻠﻜﻢ‬ omission (ḫalaw)

(cont.)
(C-1 relative to
ʿUt ̠mān)
The text of ʿUt ̠mān here has several parallels: Kor 24, 34 (allad̠īna ḫalaw min
qablikum); 10, 102 (allad̠īna ḫalaw min qablihim), and 33, 38 and 33, 62 (fī
llad̠īna ḫalaw min qablu).
However, the C-1 version is much more richly attested. We have the phrase
allad̠īna min qablikum in Kor 2, 183; 4, 26; 9, 69; 14, 9; allad̠īna min qablihim
in Kor 2, 118; 6, 148; 9, 70; 10, 12; 12, 109; 13, 42; 16, 26; 16, 33; 16, 35;
24, 55; 24, 59; 29, 3; 30, 9; 34, 45; 35, 25; 35, 44; 39, 25; 39, 50; 40, 82; 47,
10; 51, 52; 58, 5; 59, 2; 67, 18; allad̠īna min qablinā in 2, 286; allad̠īna min
qablu in 30, 42; and allad̠īna min qablika in 39, 65 and 42, 3.
2, 217 ‫ﻭﻛﻔﺮ ﺑﻪ‬ omission (wa-kufrun
bihi)
ʿUt ̠mān’s kufrun bihi is the only instance of this phrase in the entire Qurʾān.
So, one cannot consider it as expected in context.
2, 217 ‫ﻳﺮﺩﻭﻛﻢ ﻋﻦ ﺩﻳﻨﻜﻢ‬ ‫ﻳﺮﺩﻭﻛﻢ‬ omission (ʿan
dīnikum)
ʿUt ̠mān has the specifying phrase ʿan dīnikum (“from your religion”), but C-1
does not. Elsewhere in the Qurʾān, the verb yaruddūkum (“make you turn”)
and its cognates may occur with or without the specifying phrase “from your/
their/his religion”. The specifying phrase is included, for example, in both C-1
and ʿUt ̠mān in Kor 5.54 as yartadda minkum ʿan dīnikum (cf. Kor 2, 217).
However, when the verb occurs without the specifying phrase, as in 2, 109, 3,
100 and 3, 149, the context makes it clear what the person is returning from.
In other words, at first sight C-1 appears anomalous at the present point for
not specifying what one would turn away from. However, the presence in
C-1, later in the verse, of yartadid minkum ʿan dīnihi as established by the
partly visible nūn of dīnihi, can be said to remove ambiguity.
‫ﺍﻥ ﺍﻟﺬﻳﻦ ﺍﻣﻨﻮﺍ ﻭﺍﻟﺬﻳﻦ‬ ‫ﺍﻥ ﺍﻟﺬﻳﻦ ﺍﻣﻨﻮﺍ ﻭ ﻫﺠﺮﻭﺍ‬
‫ﻫﺎﺟﺮﻭﺍ ﻭﺟﻬﺪﻭﺍ‬
2, 218 omission in a list
(wa-llad̠īna . . .
wa-ǧāhadū)

(cont.)
(C-1 relative to
ʿUt ̠mān)
ʿUt ̠mān has inna llad̠īna āmanū wa-llad̠īna hāǧarū wa-ǧāhadū, whereas C-1
gives inna llad̠īna āmanū wa-hāǧarū. Essentially, ʿUt ̠mān has an item in the
list that C-1 lacks, namely ǧāhadū. The other elements, found in both texts,
are āmanū and hāǧarū.
Elsewhere in the standard Qurʾān, hāǧarū never appears with āmanū unless
ǧāhadū is also present. Therefore, their juxtaposition of the pair in C-1 cannot
be labeled expected.
As for ʿUt ̠mān’s version, we find the three verbs of its list are juxtaposed also
at three other spots, in Kor 8, 72; 8, 74 and 9, 20, as llad̠īna āmanū wa-hāǧarū
wa-ǧāhadū (cf. Kor 8, 75). Therefore, the presence of ǧāhadū at the present
spot would not be unexpected.
2, 221 ‫ﺍﻟﻤﻐﻔﺮﺓ ﺑﺈﺫﻧﻪ‬ ‫ﺍﻟﻤﻐﻔﺮﺓ‬ omission (bi-id̠nihi)
ʿUt ̠mān’s bi-id̠nihi (“by his leave”) occurs also in eight other spots in the
Qurʾān (2, 213; 2, 255; 3, 152; 5, 16; 11, 105; 22, 65; 33, 46; 42, 51), and
bi-id̠ni (“by the leave of ”) occurs twenty-three times followed by llāh or rabb.
In none of these cases is the term mentioned in conjunction with forgiveness
(maġfira). Nonetheless, the verse at hand finds a parallel in Kor 33, 46 (wa-
dāʿiyan ilā llāhi bi-id̠nihi). Due to this parallel, we may say that the version of
ʿUt ̠mān would not be entirely unexpected.
‫ﻓﺎﻋﺘﺰﻟﻮﺍ ﺍﻟﻨﺴﺎﺀ ﻓﻲ‬ (‫)ﻓﻼ ﺗﻘﺮﺑﻮﺍ‬
‫ﺍﻟﻤﺤﻴﺾ ﻭﻻ ﺗﻘﺮﺑﻮﻫﻦ‬ (‫)ﻣﺤﻴﻀﻬﻦ‬ ‫ﺍﻟﻨﺴﺎﺀ ﻓﻲ‬
2, 222 omission ( fa-ʿtazilū)
‫ﺣﺘﻰ ﻳﻄﻬﺮﻥ‬ ‫ﺣﺘﻰ ﻳﺘﻄﻬﺮﻥ‬

in addition to word
order & minor
elements.
The key difference here is the presence of the imperative fa-ʿtazilū in ʿUt ̠mān
and its absence in C-1. (The other differences are reducible to phonetic
conservation and minor elements.) The verb iʿtazala occurs at six other points
in the Qurʾān, including only once as an imperative, but never in a marital or
sexual context. There are absolutely no parallels for the verse at hand. Therefore,
the verb fa-ʿtazilū is not something that a scribe would have anticipated here.
ʿUt ̠mān’s version is not a candidate for auto-contamination. See Kor 4, 90; 4,
91; 18, 16; 19, 48; 19, 49; 44, 21.

(cont.)

(C-1 relative to
ʿUt ̠mān)
‫ﻓﺈﻥ ﺟﺎﺀﻭﻙ ﻓﺎﺣﻜﻢ‬ ‫ﻓﺎﺣﻜﻢ ﺑﻴﻨﻬﻢ‬

‫ﺑﻴﻨﻬﻢ‬
5, 42 omission (fa-in
ǧāʾūka)
The phrase in ʿUt ̠mān is not attested elsewhere. The word ǧāʾūka appears four
other times in the Qurʾān (Kor 4, 62; 4, 64; 6, 25; 58, 8), twice preceded by
id̠ā, but nowhere by in. The word ǧāʾaka occurs eleven times in the standard
6, 54; 10, 94; 13, 37; 60, 12; 63, 1; 80, 8. The word ‫ ﺟﺎﺀﻭﻙ‬also occurs in C-1
Qurʾān, preceded three times by id̠ā. See 2, 120; 2, 145; 3, 61; 5, 48; 6, 34;
at Kor 63, 4, where it harks back to the ǧāʾaka at the beginning of that sūra
(see below).
5, 44 ‫ﻓﻼ ﺗﺨﺸﻮﺍ ﺍﻟﻨﺎﺱ‬ ‫ ﻓﻼ ﺗﺨﺸﻮﻫﻢ‬omission (l-nās)
ʿUt ̠mān’s phrase fa-lā taḫšawu l-nāsa occurs nowhere else in the Qurʾān,
although it has parallels in Kor 4, 77 (farīqun minhum yaḫšawna l-nāsa ka-
ḫašyati llāhi) and 33, 37 (wa-taḫšā l-nāsa wa-llāhu aḥ aqqu an taḫšāhu). The
word al-nās is a common word in the Qurʾān, occurring 247 times.
On the other hand, C-1’s phrase fa-lā taḫšawhum is attested in Kor 2, 150
and 5, 3.
5, 49 ‫ﻓﺎﻋﻠﻢ ﺃﻧﻤﺎ‬ (‫)ﻓﺈﻧﻤﺎ‬ omission (fa-ʿlam)
ʿUt ̠mān has fa-in tawallaw fa-ʿlam annamā yurīdu llāhu an yuṣībahum bi-baʿḍi
d̠unūbihim. ʿUt ̠mān’s juxtaposition of the verbs tawallaw (in a conditional
clause) and fa-ʿlam also occurs in Kor 5, 92 (fa-in tawallaytum fa-ʿlamū), Kor
8, 40 (wa-in tawallaw fa-ʿlamū), Kor 9, 3 (wa-in tawallaytum fa-ʿlamū)
On the other hand, the version of C-1 has even more parallels: Kor 2, 137
(wa-in tawallaw fa-innamā), 3, 20 (same), 16, 82 (same, with fa), 24, 54
(same, with fa), 64, 12 (fa-in tawallaytum fa-innamā).
63, 1 ‫ﺍﻟﻤﻨﻔﻘﻮﻥ ﻗﺎﻟﻮﺍ‬ ‫ﺍﻟﻤﻨﻔﻘﻮﻥ‬ omission (qālū)
The word qālū is ubiquitous in the Qurʾān.

(cont.)
(b) “Additions” in C-1 (i.e. pluses of C-1).
2, 193 ‫ﻭﻳﻜﻮﻥ ﺍﻟﺪﻳﻦ ﷲ‬ ‫ﻭﻳﻜﻮﻥ ﺍﻟﺪﻳﻦ ﻛﻠﻪ ﷲ‬ addition (kulluhu)

The entire sentence in C-1 occurs elsewhere in ʿUt ̠mān, in Kor 8, 39. So, the
word kulluhu would have been expected here for somebody familiar with the
language of the Qurʾān.
ʿUt ̠mān’s version, on the other hand, appears nowhere else.
2, 200 ‫ﺍﺗﻨﺎ ﰲ )ﻫﺬﻩ ﺍﳊﻴﻮﺓ( ﺍﻟﺪﻧﻴﺎ ﺍﺗﻨﺎ ﰲ ﺍﻟﺪﻧﻴﺎ‬ addition (hādhihi

l-ḥ ayāt)
C-1’s version is a candidate for the assimilation of parallels at Kor 3, 117; 7,
‫ ﻭﺍﻟﺬﻳﻦ ﻫﺪﻭﺍ ﻳﺤﻜﻤﻮﻥ ﻟﻠﺬﻳﻦ ﻫﺎﺩﻭﺍ ﻭﺍﻟﺮﺑﻨﻴﻮﻥ‬addition (yaḥ kumūna

156; 11, 60; 16, 30; 20, 72; 28, 42; 29, 64; 39, 10; and 40, 39.
‫ﻭﺍﻷﺣﺒﺎﺭ ﺑﻤﺎ ﺍﺳﺘﺤﻔﻈﻮﺍ‬ ((‫ ﺑﻤﺎ ﻧﺰﻝ ﺍﷲ ﻓﻴﻬﺎ ))ﻭ‬bi-mā nazzala llāhu
5, 44
‫ﻣﻦ ﻛﺘﺐ ﺍﷲ ﻭﻛﺎﻧﻮﺍ ﻋﻠﻴﻪ‬ ((‫ ﻳﺤﻜﻢ ﺑﻪ ﺍﻟ))ﺭﺑﻨﻴ‬fīhā and yaḥ kumu
‫ﺷﻬﺪﺍﺀ‬ ‫ ﻭﻥ ﻭ ﺍﻻ))ﺣﺒﺮ(( ﺑﻤﺎ‬bihi)
‫ﺍﺳﺘﺤﻔﻈﻮﺍ ﻣﻦ ﻛﺘﺐ ﺍﷲ‬
‫))ﻭ(( ﻛﺎﻧﻮﺍ ﻋﻠﻴﻪ ﺷﻬﺪﺍﺀ‬
C-1’s yaḥ kumūna bi-mā nazzala (read anzala?) llāhu fīhā is closely paralleled
by another phrase in the same verse: man lam yaḥ kum bi-mā anzala llāhu, a
phrase that repeats two more times in the next two verses. Then in the verse
after those (5, 48) we have fa-ḥ kum baynahum fī-mā anzala llāhu. A similar
phrase appears in the next verse as well (5, 49). Therefore, the additional terms
in C-1 mirror language that repeats five times in close proximity to the spot
in question, and therefore would not be “unexpected”. It is a candidate for
assimilation of nearby phrases.
5, 45 ‫ﻭ ﻛﺘﺒﻨﺎ ﻋﻠﻴﻬﻢ‬ ‫ﻭ ﻛﺘﺒﻨﺎ ﻋﻠﻰ ﺑﻨﻲ ﺍﺳﺮﺍﺋﻞ‬ addition (ʿalā banī
Isrāʾīla)
The phrase katabnā ʿalā banī Isrāʾīla occurs in Kor 5, 32, i.e. shortly before the
verse at hand. The proximity of this occurrence makes it easy to see how the
addition in C-1 could be a false memory harking back to the earlier use.

(cont.)
63, 3 ‫ﺫﻟﻚ ﺑﺄﻧﻬﻢ ﺀﺍﻣﻨﻮﺍ ﺛﻢ‬ ‫ﺫﻟﻚ ﺑﺄﻧﻬﻢ ﺀﺍﻣﻨﻮﺍ ﺛﻢ‬ addition (t̠umma
‫ﻛﻔﺮﻭﺍ‬ ‫ﻛﻔﺮﻭﺍ‬ zdādū kufran)
‫)ﺛ ( ﻡ ﺍ) ﺯ( ﺩﺩﻭ )ﺍ ﻛﻔﺮ (ﺍ‬
The version in C-1 is a restatement of Kor 4, 137 (t̠umma āmanū t̠umma
kafarū t̠umma zdādū kufran). Cf. 3, 90 and 3, 178.
63, 3 ‫ ﻓﻄﺒﻊ ﻋﻠﻰ‬. . . ‫ﺫﻟﻚ ﺑﺄﻧﻬﻢ‬ ‫ ﻓﻄﺒﻊ ﻋﻠﻰ‬. . . ‫ﺫﻟﻚ ﺑﺄﻧﻬﻢ‬ addition (d̠ālika bi-
‫ﻗﻠﻮﺑﻬﻢ ﻓﻬﻢ ﻻ ﻳﻔﻘﻬﻮﻥ‬ ‫ﻗﻠﻮﺑﻬﻢ‬ annahum qawmun)
‫ﺫﻟﻚ ﺑﺄﻧﻬﻢ ﻗﻮﻡ ﻻ ﻳﻔﻘﻬﻮﻥ‬
C-1’s phrase d̠ālika bi-annahum is simply a repetition of the same term
earlier in the verse. The construct qawmun lā occurs seven times (Kor 5, 58;
8, 65; 9, 6; 9, 127; 43, 88; 59, 13; 59, 14), including three times as qawmun
lā yafqahūna (Kor 8, 65; 9, 127; 59, 13).
On the other hand, ʿUt ̠mān’s phrase fa-hum lā yafqahūna is also attested; see
Kor 9, 87. There are also other instances where lā yafqahūna is used without a
preceding qawm, viz. Kor 7, 179; 48, 15; 63, 7.
63, 7 ‫ﻳﻨﻔﻀﻮﺍ‬ ‫ﻳﻨﻔﻀﻮﺍ ﻣﻦ ﺣﻮﻟﻪ‬ addition (min
ḥ awlihi)
The version of C-1 has a parallel in Kor 3, 159 (la-nfaḍḍū min ḥ awlika).
63, 8 ‫ﻭﷲ ﺍﻟﻌﺰﺓ‬ ‫ﻭﷲ ﺍﻟﻌﺰﺓ ﺟﻤﻴﻌﺎ‬ addition (ǧamīʿan).
C-1’s version has the word ǧamīʿan, which occurs in the Qurʾān forty-nine
times. For close parallels of C-1’s expression, see 35, 10 (fa-li-llāhi l-ʿizzatu
ǧamīʿan); 4, 139 ( fa-inna l-ʿizzata li-llāhi ǧamīʿan); 10, 65 (inna l-ʿizzata
li-llāhi ǧamīʿan); 13, 31 (li-llāhi l-amru ǧamīʿan); 39, 44 (li-llāhi l-šafāʿatu
ǧamīʿan); 2, 165 (anna l-quwwata li-llāhi ǧamīʿan).
As for ʿUt ̠mān’s phrase wa-li-llāhi l-ʿizzatu, nowhere else does it occur in this
form; elsewhere it appears only with ǧamīʿan as in C-1’s version here.
63, 11 ‫ﺇﺫﺍ‬ ‫)ﻟﻐﺪ( ﺇﻥ‬ addition (li-ġadin)

(cont.)
The C-1 word li-ġadin is attested in Kor 59, 18. The contexts are linguistically
similar in that both spots are followed with the phrase wa-llāhu ḫabīrun bi-
mā taʿmalūn, an expression that occurs in the Qurʾān in only five other spots.
Note also the relative probable proximity of this attestation to the spot at
hand. (Incidentally, the word ġadan occurs in the Qurʾān also in Kor 12, 12;
18, 23; 31, 34; 54, 26; and 59, 18.)
62, 6 ‫ﺃﻧﻜﻢ ﺃﻭﻟﻴﺎﺀ ﷲ ﻓﻲ ﺃﻻﺧﺮﺓ ﺃﻧﻜﻢ ﺃﻭﻟﻴﺎﺀ ﷲ‬ addition (fī l-āḫirati)
C-1 has fī l-āḫirati (“in the next world”), but ʿUt ̠mān does not. The present
verse and the next are close parallels of the pair of verses 94 and 95 in sūra 2.
Much of the wording is almost exactly the same. One key difference is that
Kor 2, 94 mentions the next world (l-dāru l-āḫiratu), but the present verse
does not. Therefore, C-1’s fī l-āḫirati possibly harks back to Kor 2, 94-5. Cf.
the parallel 3, 142-3, which also concerns the next world but does not employ
the term āḫira.
The phrase fī l-āḫirati occurs twenty-four times in the Qurʾān, and āḫira
occurs 113 times.
(c) Substitutions without phonetic conservation.

‫ﻓﻤﻦ ﻛﺎﻥ ﻣﻨﻜﻢ ﻣﺮﻳﻀﺎ‬ ‫ﻓﺎﻥ ﻛﺎﻥ ﺃﺣﺪ ﻣﻨﻜﻢ‬
‫ﻣﺮﻳﻀﺎ‬
2, 196 substitution
(syntactical structure:
in . . . aḥ ad vs.
man) and phonetic
conservation (same
three words used in
different syntactical
structures).

(cont.)
The syntactical structures used by C-1 and ʿUt ̠mān occur elsewhere in the
Qurʾān.
On the one hand, the language used by C-1 is quite familiar. The C-1 phrase
aḥ adun minkum and the construct in kāna(t) are both attested in legal passages
in ʿUt ̠mān: The C-1 construct in kāna(t) occurs about forty times in ʿUt ̠mān,
including a dozen times as part of a legal prescription: Kor 2, 280; 2, 282; 4, 11
(three times), 4, 12 (three times), 4, 92 (twice), and 4, 102. The phrase aḥ adun
min occurs three times in ʿUt ̠mān, twice in the form of aḥ adun minkum. The
occurrences are all by way of legal prescription (Kor 43, 29; 5, 6; 9, 6).
On the other hand, ʿUt ̠mān’s version of this verse, with man kāna minkum
marīḍan, too, has several close parallels as part of a legal prescription: Kor 2,
185; 2, 196; 2, 232; 4, 6 (twice). The sequence man kāna occurs thirty-seven
times in ʿUt ̠mān, usually in a non-legal passage.
For another instance of C-1’s propensity to construct conditionals with in

instead of man, see below, Kor 5, 45.
2, 197 ‫ﺗﻔﻌﻠﻮﺍ‬ ‫ﺗﻌﻤﻠﻮﺍ‬ substitution (taf ʿalū
vs. taʿmalū)
The two words are graphically similar, so the difference might be due to
copying from manuscript.
ʿUt ̠mān’s mā taf ʿalū min ḫayrin is attested also in Kor 2, 215 and 4, 127.
The synonyms are both very common verbs in the Qurʾān. The verb faʿala of
ʿUt ̠mān occurs in the Qurʾān 108 times, while the verb ʿamala of C-1 occurs
some 359 times in different conjugations.
2, 209 ‫ﺟﺎﺋﺘﻜﻢ ﺍﻟﺒﻴﻨﺖ‬ ‫ﺟﺎﺀﻛﻢ ﺍﻟﻬﺪﻯ‬ substitution
(ǧāʾatkumu l-bayyināt
vs. ǧāʾakumu l-hudā)
Both phrases occur elsewhere. We have ǧāʾahumu l-hudā in Kor 17, 94 and 18,
55 (cf. 28, 37 and 28, 85). There is likewise ǧāʾakum bayyinatun in Kor 6, 157;
7, 73; and 7, 85 (cf. 40, 28).

(cont.)
2, 213 ‫ﻓﺒﻌﺚ‬ ‫ﻓﺄﺭﺳﻞ‬ substitution:

synonyms (fa-baʿat̠a
vs. fa-arsala)
The verb baʿat̠a occurs sixty-five times in the Qurʾān, mostly in connection
with resurrection. The verb arsala occurs 172 times. While both words can be
used for the sending of prophets (i.e. the subject of the verse at hand), C-1’s
arsala is used significantly more frequently in such a context.
‫ﻧﻮﺭ ﻭﻣﺼﺪﻗﺎ ﻟﻤﺎ ﺑﻴﻦ ﻳﺪﻳﻪ‬ ‫ﻧﻮﺭ ﻭ ﻣﺼﺪﻗﺎ ﻟﻤﺎ ﺃﻧﺰﻟﻨﺎ‬
‫ﻣﻦ ﺍﻟﺘﻮﺭﻳﺔ‬ ‫ﻣﻦ ﺍﻟﺘﻮﺭﻳﺔ‬
5, 46 substitution (bayna
yadayhi vs. anzalnā)
ʿUt ̠mān’s expression bayna yaday (or bayna aydī) occurs forty times in the
Qurʾān. Close parallels would include Kor 2, 97; 3, 3; 3, 50; 5, 46 (another
instance), 5, 48; 6, 92; 10, 37; 12, 111; 35, 31; 36, 9; 46, 30, and 61, 6.
As for the C-1 expression, the verb anzala is juxtaposed with al-tawrāt in Kor
3, 3; 3, 65, and 5, 44, in the form of anzalnā in the first and last cases. The
verb anzalnā occurs fifty-five times in the Qurʾān.
In sum, the expressions are both attested elsewhere, and are both attested in
close proximity to the point at hand.
5, 46 ‫ﻟﻠﻤﺘﻘﻴﻦ‬ ‫ﻟﻘﻮﻡ )ﻳﻮﻣﻨﻮﻥ‬
substitution (li-l-
(‫ ﻳﻮﻗﻨﻮﻥ‬qawmin yuʾminūna
muttaqīna vs. li-
or
or li-qawmin
yūqinūna).
ʿUt ̠mān has mawʿiẓatan li-l-muttaqīna, a phrase that occurs in three other
spots, viz. Kor 2, 66; 3, 138; and 24, 34. (The word li-l-muttaqīna occurs
seventeen other times in the Qurʾān.)
The wording in C-1 is mawʿiẓatan li-qawmin yuʾminūna, or, possibly, li-

qawmin yūqinūna. Elsewhere, the word mawʿiẓatan is never followed by such
a phrase. The phrase li-qawmin yuʾminūna occurs in thirteen spots: Kor 6,
99; 7, 52; 7, 188; 7, 203; 12, 111; 16, 64; 16, 79; 27, 86; 28, 3; 29, 24; 29,
51; 30, 37; 39, 52. The word yūqinūna occurs eleven times, including four
times within the phrase li-qawmin yūqinūna: Kor 2, 118; 5, 50; 45, 4; 45, 20.
Note that the instance at Kor 5, 50 is close to the spot at hand and one could
imagine it having influenced the phrasing of the verse at hand.

(cont.)
5, 49 ‫ﺃﻧﺰﻝ ﺍﷲ‬ ‫ﺃﻭﺣﻰ ﺍﷲ‬ substitution

(synonyms anzala vs.
awḥ ā)
The verb in ʿUt ̠mān, anzala, occurs 190 times in different senses and
conjugations, including twenty-nine times as anzala llāhu. The term is used
not only with revelations, but also in other contexts.
The verb in C-1, awḥ ā, occurs seventy-three times in various conjugations,

including eight times as awḥ ā. The term is used specifically with revelation.
Neither term could be considered unexpected in context.

63, 4 ‫ﻭﺇﺫﺍ ﺭﺃﻳﺘﻬﻢ‬ (‫ﻭﺇﺫﺍ )ﺟﺎﺅﻙ‬ substitution
(raʾaytahum instead
of ǧāʾūka)
C-1’s ‫ ﺟﺎﺀﻭﻙ‬harks back to ‫ ﺟﺎﺀﻙ‬in the first verse of the sūra. The word ǧāʾaka
occurs eleven times in the Qurʾān, including three times following id̠ā. See
above, Kor 5, 42.
ʿUt ̠mān’s raʾaytahum occurs elsewhere three times (Kor 20, 92; 33, 19; 76, 19),
but never in a similar context. The closest parallel would be Kor 76, 19.
(d) Errors of the Hand.

2, 196 ‫ﻓﻔﺪﻳﺔ‬ ‫ﻓﺪﻳﺔ‬ minor omission (fa)
probable error of the
‫ﻓﺈﺫﺍ ﻗﻀﻴﺘﻢ‬ ‫ﻭﺇﺫﺍ ﺃﻗﻀﻴﺘﻢ‬

hand
2, 200 scribal error;
‫ﻛﺴﺒﻮﺍ‬ ‫ﻛﺴﺒﻮﻥ‬
duplication of alif
2, 202 scribal error: nūn
instead of alif
For the same type of
error, see below, Kor
2, 217.

(cont.)

2, 213 ‫ﻟﻴﺤﻜﻢ ﺑﻴﻦ ﺍﻟﻨﺎﺱ‬ ‫ﻟﻴﺤﻜﻢ )ﺃ(ﺑﻴﻦ ﺍﻟﻨﺎﺱ‬ scribal error: addition
‫ﺇﻥ ﺍﺳﺘﻄﻌﻮﺍ‬ (‫ﺇﻥ ﺍﺳﺘﻄﻌﻮ )ﻥ‬

of alif
2, 217 scribal error: nūn
instead of alif
For the same type of
error, see above, Kor
‫ﻭﻟﻜﻦ‬ ‫ﻭﻟﻜﻦ )ﺇ(ﻥ‬

2, 202.
63, 7 scribal error
(e) Illegible terms that differ from ʿUt ̠mān.

5, 54 ‫ﺃﺫﻟﺔ‬ ‫؟؟؟‬ Illegible
5, 54 ‫ﺃﻋﺰﺓ‬ ‫؟؟؟‬ Illegible
15, 68 ‫ﺇﻥ ﻫﺆﻻﺀ‬ ‫؟‬ Illegible
89, 27 ‫ﺍﻟﻨﻔﺲ ﺍﻟﻤﻄﻤﺌﻨﺔ‬ (-‫ﻤ‬-) ‫ﺍﻟﻨﻔﺲ‬ illegible; error of the
‫ﺍﻟﻤﻄﻤﺌﻨﺔ‬ hand?
89, 28 ‫ﺍﺭﺟﻌﻲ ﺇﻟﻰ‬ ‫؟‬- ‫ﺍ‬ illegible (there is
room for only one or
two more letters after
the alif )
Appendix 3: Sūra Titles in C-1.
Verse C-1 Note

63 End ‫ﻫﺬﻩ ﺧﺘﻤﺔ ﺳ)ﻭﺭﺓ( ﺍﻟﻤﻨﻔﻘﻴﻦ‬ sūra title
62 End (‫)ﻫﺬﻩ( ﺧﺘﻤﺔ ﺳﻮﺭﺓ ﺍﻟﺠﻤ)ﻌﺓ‬ sūra title
89 End ‫ﮬ )ﺫﻩ( ﺧﺘﻤﺔ ﺍﻟ )ﻓﺠﺮ( ﻭﻟﻴﻞ ﺍﻟﻌﺸﺮ‬ sūra title

Appendix 4: Orthographical Characteristics of C-1
have good reason for the readings given here. Although dots and hamzas are
used below, they tend not to appear in the original.
Verse (Modern Saʿūdī C-1

Qurʾān)
2, 193 ‫ﺣﺘﻰ‬ ‫ﺣﺘﺎ‬
2, 194 ‫ ﺍﻟﺤﺮﺍﻡ‬. . . ‫ﺍﻟﺤﺮﺍﻡ‬ ‫ ﺍﻟﺤﺮﻡ‬. . . ‫ﺍﻟﺤﺮﻡ‬
2, 194 ‫ﻗﺼﺎﺹ‬ ‫ﻗﺼﺺ‬
2, 196 ‫ﺣﺘﻰ‬ ‫ﺣﺘﺎ‬
2, 196 ‫ﻣﻦ ﺻﻴﺎﻡ‬ ‫ﻣﻦ ﺻﻴﻢ‬
2, 196 ‫ﻓﺼﻴﺎﻡ‬ ‫ﻓﺼﻴﻢ‬
2, 196 ‫ﻛﺎﻣﻠﺔ‬ ‫ﻛﻤﻠﺔ‬
2, 196 ‫ﺣﺎﺿﺮﻯ‬ ‫ﺣﻀﺮﻯ‬
2, 196 ‫ﺍﻟﺤﺮﺍﻡ‬ ‫ﺍﻟﺤﺮﻡ‬
2, 196 ‫ﺟﺪﺍﻝ‬ ‫ﺟﺪﻝ‬
2, 197 ‫ﺍﻟﺰﺍﺩ‬ ‫ﺍﻟﺰﺩ‬
2, 197 ‫ﻳﺄﻭﻟﻰ ﺍﻷﻟﺒﺐ‬ ‫ﻳﺄﻭﻻ ﺍﻷﻟﺒﺐ‬
2, 198 ‫ﺟﻨﺎﺝ‬ ‫ﺟﻨﺢ‬
2, 198 ‫ﺍﻟﺤﺮﺍﻡ‬ ‫ﺍﻟﺤﺮﻡ‬
2, 198 ‫ﺍﻟﻀﺎﻟﻴﻦ‬ ‫ﺍﻟﻀﻠﻴﻦ‬
2, 199 ‫ﺃﻓﺎﺽ‬ ‫ﺃﻓﺾ‬
2, 203 ‫ﺃﻳﺎﻡ‬ ‫ﺃﻳﻢ‬
2, 204 ‫ﺍﻟﺨﺼﺎﻡ‬ ‫ﺍﻟﺨﺼﻢ‬
2, 206 ‫ﺍﻟﻤﻬﺎﺩ‬ ‫ﺍﻟﻤﻬﺪ‬
2, 207 ‫ﻣﺮﺿﺎﺕ‬ (‫)ﻣﺮﺿﺖ‬
2, 207 ‫ﺑﺎﻟﻌﺒﺎﺩ‬ ‫ﺑﺎﻟﻌﺒﺪ‬
2, 210 ‫ﺍﻟﻐﻤﺎﻡ‬ ‫ﺍﻟﻐﻤﻢ‬
2, 210 ‫ﺃﺗﻴﻨﻬﻢ‬ ‫ﺃﺗﻴﻨﻴﻬﻢ‬
(the long ā
designated with
a tooth, no dots)

‫‪B. Sadeghi, U. Bergmann / Arabica 57 (2010) 343-436‬‬ ‫‪435‬‬
‫)‪(cont.‬‬
‫‪Verse‬‬ ‫‪(Modern Saʿūdī‬‬ ‫‪C-1‬‬
‫)‪Qurʾān‬‬
‫‪2, 211‬‬ ‫ﺍﻟﻌﻘﺎﺏ‬ ‫ﺍﻟﻌﻘﺐ‬

‫‪2, 212‬‬ ‫ﺣﺴﺎﺏ‬ ‫)ﺣﺴﺐ(‬
‫‪2, 214‬‬ ‫ﺍﻟﺒﺎﺳﺎ‬ ‫ﺍﻟﺒﺴﺎ‬
‫‪2, 216‬‬ ‫ﺍﻟﻘﺘﺎﻝ‬ ‫ﺍﻟﻘﺘﻞ‬
‫‪2, 217‬‬ ‫ﺍﻟﺤﺮﺍﻡ‬ ‫ﺍﻟﺤﺮﻡ‬
‫‪2, 217‬‬ ‫ﻗﺘﺎﻝ‬ ‫ﻗﺘﻞ‬
‫ﻗﺘﺎﻝ‬ ‫ﻗﺘﻞ‬
‫ﻻ ﻳﺰﺍﻟﻮﻥ‬ ‫ﻻ ﻳﺰﻟﻮﻥ‬
‫‪2, 217‬‬
‫‪2, 217‬‬
‫)‪(verso‬‬
‫‪2, 217‬‬ ‫ﻳﻘﺎﺗﻠﻮﻧﻜﻢ‬ ‫ﻳﻘﺘﻠﻮﻧﻜﻢ‬
‫‪2, 218‬‬ ‫ﻫﺎﺟﺮﻭﺍ‬ ‫ﻫﺠﺮﻭﺍ‬
‫‪2, 218‬‬ ‫ﺭﺣﻤﺖ‬ ‫ﺭﺣﻤﺔ‬
‫‪2, 219‬‬ ‫ﻳﺴﺌﻠﻮﻧﻚ‬ ‫ﻳﺴﺄﻟﻮﻧﻚ‬
‫‪2, 219‬‬ ‫ﻭﻳﺴﺌﻠﻮﻧﻚ‬ ‫ﻭﻳﺴﺄﻟﻮﻧﻚ‬
‫‪2, 220‬‬ ‫ﺇﺻﻼﺡ‬ ‫ﺇﺻﻠﺢ‬
‫‪2, 220‬‬ ‫ﺗﺨﺎﻟﻄﻮﻫﻢ‬ ‫ﺗﺨﻠﻄﻮﻫﻢ‬
‫‪5, 44‬‬ ‫ﻫﺎﺩﻭﺍ‬ ‫ﻫﺪﻭﺍ‬
‫‪5, 44‬‬ ‫ﺑﺄﻳﺘﻲ‬ ‫ﺑﺄﻳﺌﺘﻲ‬
‫‪without diacritics‬‬
‫)‪(just extra tooth‬‬
‫‪5, 48‬‬ ‫ﻋﻤﺎ‬ ‫ﻋﻦ ﻣﺎ‬
‫‪5, 48‬‬ ‫ﻣﻨﻬﺎﺟﺎ‬ ‫ﻣﻨﻬﺠﺎ‬
‫‪5, 48‬‬ ‫ﻓﻲ ﻣﺎ‬ ‫ﻓﻴﻤﺎ‬
‫‪5, 52‬‬ ‫ﺩﺍﺋﺮﺓ‬ ‫ﺩﺋﺮﺓ‬
‫‪15, 55‬‬ ‫ﺍﻟﻀﺎﻟﻮﻥ‬ ‫ﺍﻟﻀﻠﻮﻥ‬
‫‪15, 58‬‬ ‫ﻗﺎﻟﻮﺍ‬ ‫ﻗﻠﻮﺍ‬
‫‪15, 71‬‬ ‫ﻗﺎﻝ‬ ‫ﻗﻞ‬
‫‪63, 4‬‬ ‫ﺃﺟﺴﺎﻣﻬﻢ‬ ‫ﺃﺟﺴﻤﻬﻢ‬
‫‪Downloaded from Brill.com12/28/2021 08:48:21AM‬‬

‫‪via Fondation du Roi Abdul Aziz‬‬
‫‪436‬‬ ‫‪B. Sadeghi, U. Bergmann / Arabica 57 (2010) 343-436‬‬
‫)‪(cont.‬‬
‫‪Verse‬‬ ‫‪(Modern Saʿūdī‬‬ ‫‪C-1‬‬

‫)‪Qurʾān‬‬
‫‪63, 5‬‬ ‫ﺗﻌﺎﻟﻮﺍ‬ ‫)ﺗﻌﻠﻮﺍ(‬

‫‪63, 5‬‬ ‫ﺭﺃﻳﺘﻬﻢ‬ ‫ﺭﻳﺘﻬﻢ‬
‫‪63, 7‬‬ ‫ﺣﺘﻰ‬ ‫ﺣﺘﺎ‬
‫‪63, 7‬‬ ‫ﺧﺰﺁﺋﻦ‬ ‫ﺧﺰﺋﻦ‬
‫‪62, 5‬‬ ‫ﺍﻟﺤﻤﺎﺭ‬ ‫ﺍﻟﺤﻤﺮ‬
‫‪62, 5‬‬ ‫ﺃﺳﻔﺎﺭﺍ‬ ‫ﺃﺳﻔﺮﺍ‬
‫‪62, 6‬‬ ‫ﻫﺎﺩﻭﺍ‬ ‫ﻫﺪﻭﺍ‬
‫‪89, 7‬‬ ‫ﺍﻟﻌﻤﺎﺩ‬ ‫ﺍﻟﻌﻤﺪ‬
‫‪89, 8‬‬ ‫ﺍﻟﺒﻠﺪ‬ ‫)ﺍﻟﺒﻠﻴﺪ(‬
‫‪just a tooth, no‬‬
‫‪dots, for ā‬‬
‫‪89, 9‬‬ ‫ﺟﺎﺑﻮﺍ‬ ‫ﺟﺒﻮﺍ‬
‫‪89, 10‬‬ ‫ﺍﻷﻭﺗﺎﺩ‬ ‫ﺍﻷﻭﺗﺪ‬
‫‪89, 11‬‬ ‫ﺍﻟﺒﻠﺪ‬ ‫)ﺍﻟﺒﻠﻴﺪ(‬
‫‪just a tooth, no‬‬
‫‪dots, for ā‬‬
‫‪89, 14‬‬ ‫ﻟﺒﺎﻟﻤﺮﺻﺎﺩ‬ ‫ﻟﺒﺎﻟﻤﺮﺻﺪ‬
‫‪89, 18‬‬ ‫ﻃﻌﺎﻡ‬ ‫ﻃﻌﻢ‬
‫‪89, 26‬‬ ‫ﻭﺛﺎﻗﻪ‬ ‫ﻭﺛﻘﻪ‬
‫‪89, 28‬‬ ‫ﺭﺍﺿﻴﺔ‬ ‫ﺭﺿﻴﺔ‬
‫‪Downloaded from Brill.com12/28/2021 08:48:21AM‬‬

‫‪via Fondation du Roi Abdul Aziz‬‬

The Codex of A Companion of The Prophet and The Qurān of The Prophet

Uploaded by

Document Informationclick to expand document information

Document Informationclick to expand document information

Copyright:

Available Formats

The Codex of A Companion of The Prophet and The Qurān of The Prophet

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Codex of A Companion of The Prophet and The Qurān of The Prophet

Uploaded by

Copyright:

Available Formats

Arabica 57 (2010) 343-436 brill.

The Codex of a Companion of the Prophet and the

Behnam Sadeghi and Uwe Bergmann1

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

comparing a variety of Companion codices and picking the most common

Downloaded from Brill.com12/28/2021 08:48:21AM

ion, espousing a kind of codical pluralism, considered non-ʿUt ̠mānic Qurʾāns

Downloaded from Brill.com12/28/2021 08:48:21AM

Ṣanʿāʾ 1 Folios Identified and the Results of Radiocarbon Dating Described

Downloaded from Brill.com12/28/2021 08:48:21AM

Figure 1. Stanford ’07, recto.

Downloaded from Brill.com12/28/2021 08:48:21AM

Figure 2. Stanford ’07, verso.

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

For historical reasons, however, what is of greater interest is the probability

Table 1. Result of Radiocarbon Dating. The graph gives, in five-year

Probability that the

“δ13C −20.8 per mil

Downloaded from Brill.com12/28/2021 08:48:21AM

Accelerator Unit (https://c14.arch.ox.ac.uk/login/login.php?Location=/oxcal/OxCal.html). The

Downloaded from Brill.com12/28/2021 08:48:21AM

Table 2. The folios of Ṣanʿāʾ 1 studied here.

Name of Leaf and Upper Text Lower Text

Downloaded from Brill.com12/28/2021 08:48:21AM

Name of Leaf and Upper Text Lower Text

had a distinct handwriting and distinctive dotting and orthographical habits.

Downloaded from Brill.com12/28/2021 08:48:21AM

Table 3. The four hands in Ṣanʿāʾ 1, listed in the proposed chronological

Basic Description of the Folios

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

attested in several other Qurʾāns in the Ḥ iǧāzī script.23

Downloaded from Brill.com12/28/2021 08:48:21AM

Table 4. Instances where C-1 agrees with a reported non-ʿUt̠mānic

2.217 qitālin fīhi (wa)-ʿan qitālin fīhi ʿan qitālin fīhi

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

What We Know about the Codex of ʿUt̠mān

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

While the above argument concerns collective recollections of a momen-

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

The Date of the Upper Writing of Ṣanʿāʾ 1

Downloaded from Brill.com12/28/2021 08:48:21AM

The Provenance of an ʿUt̠mānic Qurʾān: Results and Methods

Downloaded from Brill.com12/28/2021 08:48:21AM

Downloaded from Brill.com12/28/2021 08:48:21AM

sion. And yet, remarkably, there are instances of a close correspondence